<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:googleplay="http://www.google.com/schemas/play-podcasts/1.0"><channel><title><![CDATA[First Order]]></title><description><![CDATA[Clear thinking on leadership, security, and the human systems that drive performance in a high-pressure world.]]></description><link>https://firstorder.zenone.org</link><image><url>https://substackcdn.com/image/fetch/$s_!kFRi!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fb3a0e95b-4eef-4aa8-9b56-feafaa0b8e66_1280x1280.png</url><title>First Order</title><link>https://firstorder.zenone.org</link></image><generator>Substack</generator><lastBuildDate>Sun, 22 Nov 2026 15:38:48 GMT</lastBuildDate><atom:link href="https://firstorder.zenone.org/feed" rel="self" type="application/rss+xml"/><copyright><![CDATA[Steve Zenone]]></copyright><language><![CDATA[en]]></language><webMaster><![CDATA[firstorder@substack.com]]></webMaster><itunes:owner><itunes:email><![CDATA[firstorder@substack.com]]></itunes:email><itunes:name><![CDATA[Steve Zenone]]></itunes:name></itunes:owner><itunes:author><![CDATA[Steve Zenone]]></itunes:author><googleplay:owner><![CDATA[firstorder@substack.com]]></googleplay:owner><googleplay:email><![CDATA[firstorder@substack.com]]></googleplay:email><googleplay:author><![CDATA[Steve Zenone]]></googleplay:author><itunes:block><![CDATA[Yes]]></itunes:block><item><title><![CDATA[From a Vendor Key to Cursor’s Extension Store]]></title><description><![CDATA[Cursor's incident shows how old AI coding traces and stolen credentials can lead somewhere a vendor agreement never intended.]]></description><link>https://firstorder.zenone.org/p/the-vendor-key-reached-the-extension</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-vendor-key-reached-the-extension</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 17 Nov 2026 17:12:02 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!fXA1!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Cursor says a key exposed by its AI vendor let an attacker copy old coding-agent records. Credentials in those records led to other Cursor systems.</p></li><li><p>The attacker later used a different stolen credential to alter Claude Code extension copies in Cursor&#8217;s download cache. Up to 5,000 users potentially received them.</p></li><li><p>I&#8217;d check retained AI data for secrets, follow what those secrets can access and verify what reaches developers&#8217; machines.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!fXA1!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!fXA1!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!fXA1!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1903687,&quot;alt&quot;:&quot;Wrapped packages on a metal shelf, with one wrapper partly open to reveal a red block.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/225264140?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Wrapped packages on a metal shelf, with one wrapper partly open to reveal a red block." title="Wrapped packages on a metal shelf, with one wrapper partly open to reveal a red block." srcset="https://substackcdn.com/image/fetch/$s_!fXA1!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!fXA1!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F95d818f9-0335-437c-8fbf-29f71d98f6ad_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">Attackers replaced copies of the Claude Code extension in Cursor's marketplace cache, without compromising the original extension.</figcaption></figure></div><p>A key issued to an AI vendor ended up being part of an attack on a software extension marketplace.</p><p>That&#8217;s the chain Cursor described on November 13, after investigating an incident that first came to light on October 29. The attacker didn&#8217;t need to compromise Cursor&#8217;s official app releases or Anthropic&#8217;s extension publisher. Cursor says the attacker used access to data from a vendor relationship, found credentials in that data and eventually put malicious copies of an extension into a cache serving downloads to users.</p><p>The distance between those two systems is what makes this incident worth examining. How did access granted for one purpose reach software being downloaded by customers?</p><h2>Start with the key and the old traces</h2><p>In <a href="https://cursor.com/blog/security-incident-update">Cursor&#8217;s November 13 incident update</a>, the company says an attacker obtained an access key Cursor had issued to Fireworks, an AI provider. Cursor says Fireworks exposed the key. The attacker used it to copy data from a read-only storage bucket used for sharing material with that provider.</p><p>The bucket held a limited set of coding-agent traces: records of interactions with a coding agent, mostly from 2025. Cursor says those sessions weren&#8217;t in Privacy Mode and it was permitted to use the material for model training. These were stored records, not live requests to AI models.</p><p>Some of the traces contained credentials that hadn&#8217;t been removed. Cursor says fewer than 0.01% of customers had API credentials exposed. It separately reports that coding-agent traces were exposed for fewer than 0.05% of users not in Privacy Mode. Not every Cursor customer had code or secrets exposed.</p><p>The attacker used credentials found in the data to reach selected online services Cursor used internally. According to the company, that access included limited support records, email conversations, product usage details, summaries of some coding-agent conversations and archived non-production source repositories. Some support records also contained credentials users had shared while seeking help.</p><p>This is where I&#8217;d examine the vendor arrangement: what the provider needed to read, why the data was retained and how credentials ended up in it. I&#8217;d also check whether a stolen credential could be used from anywhere on the internet.</p><p>I&#8217;m not saying that keeping training data automatically causes a breach. Cursor says its data sharing was permitted. The security question is whether the material was safe to retain and whether a stolen key could reach more than the vendor needed.</p><p>Fireworks has published <a href="https://fireworks.ai/security/bulletins/october-2026-security-incident-update">its own November 9 incident update</a>. It describes unauthorized activity involving internal systems and some customer data, with an investigation still underway. That account helps explain the vendor&#8217;s broader response, but it doesn&#8217;t independently reconstruct every step of the Cursor attack. I&#8217;m using Cursor&#8217;s investigation for the specific chain it reports.</p><h2>Then the attack reached a download</h2><p>On October 29, Cursor says the attacker used a different stolen credential, obtained from its internal systems, to place unauthorized copies of Anthropic&#8217;s Claude Code extension in Cursor&#8217;s extension marketplace cache, where files are stored for download.</p><p>Cursor says the malicious copies were <strong>not published by Anthropic</strong> and were <strong>not present in Open VSX</strong>, the marketplace&#8217;s upstream source. Cursor&#8217;s official app and its release-signing keys weren&#8217;t compromised either.</p><p>A customer could still request an extension through Cursor&#8217;s marketplace and receive a malicious cached copy of something they expected to be legitimate.</p><p>Cursor identified two affected versions, 2.1.312 and 2.1.313, on macOS and Windows. Its <a href="https://cursor.com/blog/claude-code-extension-security-notice">October 29 security notice</a> says the modified extension tried to send device information to the attacker, add a key for SSH remote login and leave a way to reconnect later. Cursor says attempts to enable SSH failed on most macOS and Windows hosts because the extension lacked administrator privileges. The company says the attacker-operated server is no longer functional.</p><p><a href="https://www.microsoft.com/en-us/security/blog/2026/10/30/inside-malicious-claude-code-branded-cursor-extension-persistent-ssh-backdoor/">Microsoft Security Research independently examined the injected extension code</a> and confirmed the attempted SSH key change, remote tunnel and persistence behavior.</p><p>The altered versions were available for about 11 hours. Cursor estimates that <em>at most</em> 5,000 users requesting particular cached versions could have received them. Cursor says it notified affected customers directly.</p><p>I&#8217;ve written about the need to know your suppliers and their access paths in <a href="https://firstorder.zenone.org/p/you-cant-patch-a-supply-chain-you">You Can&#8217;t Patch a Supply Chain You Can&#8217;t Inventory</a>. This incident adds an uncomfortable detail: even knowing which package you intended to download doesn&#8217;t prove the cache handed you what you expected.</p><h2>What the original permission could reach</h2><p>In <a href="https://firstorder.zenone.org/p/your-ai-workflow-is-on-the-internet">Your AI Workflow Is on the Internet. Who&#8217;s Responsible for It?</a>, I asked what a compromised automation service could reach through its connected accounts.</p><p>Here, the first key only allowed reading. The credentials in the retained data led to other systems, and a different stolen credential later allowed the attacker to change extension files. A review of the first key&#8217;s permissions wouldn&#8217;t have shown that whole route.</p><p>For a security team adopting an AI coding service, I&#8217;d ask the vendor owner and the engineering team to show three things.</p><p><strong>First, what leaves our environment and how long it stays.</strong> Privacy settings and training permissions aren&#8217;t footnotes. Determine which prompts, code, agent traces and support submissions can be stored or used by a provider. Check whether the retained records can contain access tokens, passwords or personal data. Secret scanning should apply to traces and support attachments, not only to Git repositories.</p><p><strong>Next, what a stolen vendor credential can actually read.</strong> Read-only access isn&#8217;t harmless if the data contains other working credentials. Limit each key to the data the vendor needs, restrict where it can be used when feasible and make expiration and rotation part of the arrangement. Look for credentials that have outlived the original session or project. If one turns up in a retained trace, have its owner revoke or rotate it at the service that issued it, then investigate where it was used.</p><p><strong>Finally, what customers receive from a distribution system.</strong> The download cache needed its own access controls and integrity checks. Teams should examine how cached files are verified against the publisher&#8217;s authenticated originals, who can modify them and how customers can independently verify what they&#8217;ve downloaded through cryptographic hashes, digital signatures and trusted provenance. A matching file extension and version number don&#8217;t prove it&#8217;s the right file.</p><p>Those aren&#8217;t controls that can all be delegated to the AI vendor. The provider is responsible for its own environment, the developer-tool company owns its data-sharing and distribution systems and customers still need to know which extensions and versions are permitted on their machines.</p><p>I&#8217;d also ask whether the organization has a practical way to identify affected installations and revoke exposed credentials without relying on developers to notice a blog post.</p><h2>What this incident does and doesn&#8217;t tell us</h2><p>Cursor says the vast majority of its customers weren&#8217;t affected. Prompts and code submitted in Privacy Mode weren&#8217;t part of the exposed trace data. The company also reports that its official app releases, customer code repositories and live inference requests weren&#8217;t compromised. It removed the malicious extension, rebuilt its marketplace and rotated exposed credentials, along with other internal credentials as a precaution.</p><p>Cursor says the last attacker activity it observed was on November 1, and that attempt was blocked. That&#8217;s its finding as of November 13. It doesn&#8217;t independently establish that every affected system is clean.</p><p>If you installed versions 2.1.312 or 2.1.313 of the Claude Code extension from Cursor&#8217;s marketplace on October 29, follow <a href="https://cursor.com/blog/claude-code-extension-security-notice">Cursor&#8217;s remediation instructions</a> and review <a href="https://www.microsoft.com/en-us/security/blog/2026/10/30/inside-malicious-claude-code-branded-cursor-extension-persistent-ssh-backdoor/">Microsoft&#8217;s investigation</a>. Don&#8217;t assume a machine is clean just because the suspicious SSH key isn&#8217;t there. Uninstalling the extension may leave malicious startup scripts and other persistence mechanisms behind.</p><p>I wouldn&#8217;t use this incident to argue that companies should stop using AI coding assistants or outside model providers. That&#8217;s not what the evidence shows.</p><p>What happened was more specific. Credentials that should have been redacted remained in stored data. An attacker used exposed credentials to move beyond the first compromised service, eventually reaching a system customers trusted to deliver software.</p><p>What data does the vendor receive? Where are copies stored? Who can retrieve them? And if someone steals a credential exposed through that relationship, what else can they reach?</p><p>If the security review ends at the vendor&#8217;s boundary, we&#8217;re missing part of the risk.</p><h3>Resources</h3><ul><li><p><a href="https://cursor.com/blog/security-incident-update">Cursor, An update on our recent security incident (November 13, 2026)</a></p></li><li><p><a href="https://cursor.com/blog/claude-code-extension-security-notice">Cursor, Security notice for malicious Claude Code extension versions (October 29, 2026)</a></p></li><li><p><a href="https://www.microsoft.com/en-us/security/blog/2026/10/30/inside-malicious-claude-code-branded-cursor-extension-persistent-ssh-backdoor/">Microsoft Security Research, Inside a malicious Claude Code-branded extension (October 30, 2026)</a></p></li><li><p><a href="https://fireworks.ai/security/bulletins/october-2026-security-incident-update">Fireworks, Update on October 2026 security incident (updated November 9, 2026)</a></p></li></ul>]]></content:encoded></item><item><title><![CDATA[AI Can Find the Flaws. Can We Get the Fixes Deployed?]]></title><description><![CDATA[OpenAI's Linux research produced hundreds of working exploits and candidate patches. Security teams now need to ask whether they can process findings at that speed.]]></description><link>https://firstorder.zenone.org/p/the-vulnerabilities-are-coming-faster</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-vulnerabilities-are-coming-faster</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 10 Nov 2026 16:31:07 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!Fu03!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>OpenAI says its automated Linux research reproduced 1,724 vulnerabilities and developed working local privilege-escalation exploits for 948 distinct flaws. </p></li><li><p>AI can also propose fixes. But somebody still has to verify the finding, decide what matters, get the patch accepted and ship it safely.</p></li><li><p>I&#8217;d measure the time from a confirmed, relevant finding to an effective fix, not celebrate how many findings the scanner produced.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!Fu03!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!Fu03!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!Fu03!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2032692,&quot;alt&quot;:&quot;A person examines a circuit board under a work lamp beside a conveyor of tagged boards.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/224181131?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="A person examines a circuit board under a work lamp beside a conveyor of tagged boards." title="A person examines a circuit board under a work lamp beside a conveyor of tagged boards." srcset="https://substackcdn.com/image/fetch/$s_!Fu03!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!Fu03!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0b1e1618-9d9d-4f03-b82f-b3463dc415c2_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">Finding vulnerabilities is getting faster. Verifying them and getting safe fixes into production still takes time.</figcaption></figure></div><p>On November 4, OpenAI published results from an automated security research pipeline that examined the Linux kernel. It identified 2,308 candidate findings, reproduced 1,724 vulnerabilities in a test environment and developed working local privilege-escalation exploits for 948 distinct flaws. These exploits let an unprivileged local user gain root access, or full administrative control, in the test environment. The run cost about $1.21 million in API credits.</p><p>That&#8217;s a lot of vulnerabilities to hand to somebody.</p><p>The technical achievement deserves attention. So does the part that happens after the research team produces a report. A vulnerability can be real, reproducible and serious, yet still sit between the people who found it and the people who can safely fix it. That distance is where I&#8217;d focus as a security leader.</p><h2>What the numbers actually say</h2><p>OpenAI&#8217;s researchers didn&#8217;t simply ask a model to inspect source code and count its answers. Their pipeline selected security-relevant parts of the kernel, looked for possible flaws, tried to reproduce them in isolated virtual machines and then tested whether some could be exploited. It generated candidate patches and put human researchers in the review process before reporting findings to maintainers.</p><p>That distinction is worth keeping. A candidate finding is a hypothesis. A reproducible crash gives you stronger evidence. A working exploit shows a particular security consequence under particular conditions. Maintainers still need to review a candidate patch and check whether it fixes the problem without breaking supported configurations.</p><p>OpenAI&#8217;s report says its exploit stage initially produced 1,169 successful local privilege-escalation exploits. After variant analysis, the researchers counted 948 distinct vulnerabilities and developed corresponding patches.</p><p>There&#8217;s another wrinkle in the report I appreciated. In an earlier run, 20 of 200 apparently successful exploits had gained root access through a different vulnerability than the one assigned to the agent. The team added an independent check to verify the actual exploitation path. Even when the result looks convincing, the evidence can be answering the wrong question.</p><h2>The people receiving the reports still have a job to do</h2><p>A maintainer receiving a vulnerability report has to establish whether the issue still exists on the current branch, whether someone else already reported it and whether the proposed change fixes the underlying problem without breaking something else.</p><p>OpenAI saw that work happening in real time. During the ten-day research period, 61 of the local privilege-escalation issues had already been patched upstream, 43 had staging patches and another 299 were being reported or discussed publicly. Those counts came from OpenAI&#8217;s own observations. </p><p>Mozilla offers a valuable counterexample to the idea that fixing must remain slow. In an October 29 post, Firefox engineer Bobby Holley wrote that Mozilla had shipped more than seven times as many security fixes as in the comparable period of 2025, after using frontier models to find and fix bugs. Mozilla&#8217;s October Firefox 159 advisory also names OpenAI Codex Security as the reporter for numerous high-severity issues. AI is already helping on both sides of the work.</p><p>But a candidate patch and a shipped update are different things. OpenAI says its early patch-generation attempts sometimes added unnecessary code, introduced potential performance tradeoffs or missed related variants of a vulnerability. Further instructions and iteration improved the patches, but that didn&#8217;t remove the need for review.</p><p>For a kernel maintainer, those aren&#8217;t abstract quality concerns. A bad change can affect machines far beyond the original test environment. For an enterprise security team, the chain is longer still: the upstream fix has to reach a supported distribution or vendor, then an asset owner has to test and deploy it.</p><h2>Your vulnerability backlog is not a research leaderboard</h2><p>I&#8217;ve written before about how AI reduces the human labor required to attack systems in <a href="https://firstorder.zenone.org/p/your-threat-model-still-assumes-human">Your Threat Model Still Assumes Human Attacker Labor</a>. That changes the assumptions behind some accepted risks.</p><p>I made the broader point about work waiting for review and approval in <a href="https://firstorder.zenone.org/p/the-51-point-gap-your-people-got">The 51-Point Gap: Your People Got Faster. The Work Didn't.</a> This research gives us a more specific test: is a verified vulnerability waiting on a maintainer&#8217;s review, a supported vendor update or our own deployment? OpenAI&#8217;s candidate patches and Mozilla&#8217;s shipped fixes are also a reason not to assume the repair work will stay manual. I&#8217;d want to know which part is actually holding us up.</p><p>I&#8217;d want the dashboard to tell me more than how many findings were opened and closed. How many are confirmed and reachable in our environment? How many have a fix available? How long do critical, exploitable issues sit after a usable fix exists? Which teams are waiting for a patch from a supplier, and which are waiting on our own testing or deployment?</p><p>I&#8217;d also want to separate the vulnerabilities we&#8217;re actually exposed to from the ones that are technically severe but inaccessible in our deployment. CISA&#8217;s risk-based prioritization guidance considers exposure, known exploitation, whether exploitation can be automated and technical impact. That&#8217;s a more insightful starting point than treating every high-severity label as an equal emergency.</p><p>The temptation is to buy another discovery tool because its output is easy to count. The more uncomfortable conversation is with the people who own the applications and release schedules. Do they have room to fix what we&#8217;re finding? And when a patch is risky, who can make the call to isolate the affected system, change a configuration or accept the remaining exposure?</p><h2>AI can help with the fixes, too</h2><p>I don&#8217;t think the answer is to slow vulnerability discovery. It would be strange to ask researchers to stop finding real flaws because the rest of us haven&#8217;t worked out how to handle the reports.</p><p>Automated reproduction and patch generation can remove real work from the process. OpenAI&#8217;s research demonstrates that. Mozilla&#8217;s experience suggests those gains can reach shipped software, too. The useful question is where automation can reduce the work without removing the checks that make a fix trustworthy.</p><p>That might mean rechecking a report against the latest source before it reaches a maintainer. It might mean testing whether a patch handles related variants, running regression suites or preparing enough evidence for an engineer to make a decision quickly. If an issue is already fixed upstream, don&#8217;t send maintainers another report as though it&#8217;s new. For systems we run, keep tracking it until we verify that the fix or another effective protection is in place.</p><p>There&#8217;s no single clock for all of this. A flaw that can be exploited directly through an internet-facing service isn&#8217;t the same operational problem as a local privilege-escalation flaw that first requires the attacker to run code on the affected machine. A vendor patch awaiting deployment isn&#8217;t the same as an unpatched upstream defect. An impactful security program needs to distinguish them before promising a deadline.</p><p>For the vulnerabilities that actually put us at risk, I&#8217;d measure how long it takes to get from a verified finding to a fix running in our environment. I&#8217;d keep the age of still-open findings visible, too. Otherwise, we&#8217;re only measuring the work we&#8217;ve finished.</p><p>Then I&#8217;d look at where the time went. Are we waiting for a maintainer or vendor, or for our own testing and deployment? If we can&#8217;t patch safely yet, I&#8217;d want a named owner for reducing the exposure while we wait.</p><h2>Resources</h2><ul><li><p><a href="https://openai.com/index/a-thousand-fewer-paths-to-root/">OpenAI, A thousand fewer paths to root (November 4, 2026)</a></p></li><li><p><a href="https://openai.com/index/introducing-gpt-6-cyber/">OpenAI, GPT-6 Cyber (November 5, 2026)</a></p></li><li><p><a href="https://blog.mozilla.org/en/firefox/privacy-security/defending-patch-gap/">Mozilla, Defending the patch gap against automated attackers (October 29, 2026)</a></p></li><li><p><a href="https://www.mozilla.org/en-US/security/advisories/mfsa2026-118/">Mozilla, Firefox 159 security advisory (October 27, 2026)</a></p></li><li><p><a href="https://content.govdelivery.com/accounts/USDHSCISA/bulletins/426df53">CISA, Vulnerability Review and risk-based prioritization (2026)</a></p></li></ul>]]></content:encoded></item><item><title><![CDATA[Your AI Bill Might Be a Security Incident]]></title><description><![CDATA[Mandiant found criminals were using stolen AI access at a company's expense. Who can catch it and stop the charges?]]></description><link>https://firstorder.zenone.org/p/your-ai-bill-might-be-a-security</link><guid isPermaLink="false">https://firstorder.zenone.org/p/your-ai-bill-might-be-a-security</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 03 Nov 2026 19:17:59 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!A_w9!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Mandiant investigated 2026 attacks in which criminals used stolen cloud credentials and AI keys to run models at someone else&#8217;s expense. Some incidents generated over $100,000 in reported charges in one day.</p></li><li><p>A budget alert can tell you costs are climbing without stopping them. Actual spending limits differ by provider and service.</p></li><li><p>I&#8217;d test who notices unauthorized model usage, who can revoke the access and what happens to legitimate work when they do.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!A_w9!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!A_w9!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!A_w9!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2324092,&quot;alt&quot;:&quot;A metal usage meter and orange shutoff lever on a clear tube filled with blue and amber tokens.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/223349502?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="A metal usage meter and orange shutoff lever on a clear tube filled with blue and amber tokens." title="A metal usage meter and orange shutoff lever on a clear tube filled with blue and amber tokens." srcset="https://substackcdn.com/image/fetch/$s_!A_w9!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!A_w9!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F92b0baf5-2000-4879-bcc1-b2f32af67f2e_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A spending alert reports costs. Someone still needs the access and authority to stop unauthorized usage. AI-generated editorial illustration.</figcaption></figure></div><p>On November 2, <a href="https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models/">Mandiant published details of attacks</a> that caused more than $100,000 in fraudulent AI usage charges in a single day. In some of the incidents it investigated, cumulative charges approached $1 million before the attackers lost access.</p><p>The people running up those bills weren&#8217;t buying AI services with their own money. They were using cloud credentials and model access that belonged to someone else. Mandiant calls it <em>LLMJacking</em>, meaning someone hijacks access to large language models, often to consume or resell the service.</p><p>Mandiant doesn&#8217;t say which team at each victim organization discovered the abuse, or how much the companies ultimately paid after any disputes. </p><p>Here&#8217;s the question I&#8217;d put in front of the security leader: <strong>If somebody uses a stolen credential to run up our AI bill, who has the access and authority to shut it down?</strong></p><h2>How a stolen key becomes a very expensive service</h2><p>AI models are usually billed by usage. An application sends a request to a model through an API, a connection that lets software use the model directly. The provider processes the request and records usage against the account tied to the credential. Depending on the service, that can mean paying for the text processed, the output generated or other compute resources.</p><p>The application may be completely legitimate. The problem starts when somebody else gets the credential.</p><p>In Mandiant&#8217;s investigations, the common entry points included stolen cloud login tokens, keys for service accounts used by software and model API keys accidentally exposed in public code repositories. Some attackers used those credentials to turn on AI services that the victim hadn&#8217;t been using. Others created new keys and accounts so they could keep making requests.</p><p>The attackers didn&#8217;t always use the models themselves. Mandiant describes services that collected stolen AI credentials and routed paying customers&#8217; requests through them. Someone bought cheap model access from the criminal service. The model provider billed the organization whose credential was being used.</p><p>One of those operations sent <strong>more than 500,000 model API requests in less than eight hours</strong>, according to Mandiant. The researchers attributed nearly <strong>$1 million in token costs</strong> to a victim organization in that case. That&#8217;s the cost of the model usage reported in the investigation.</p><p>The model provider can see a valid key and bill the account attached to it, even if the person using the key stole it. A security team might have excellent endpoint protection on employees&#8217; laptops and still have to answer why a cloud service account suddenly began purchasing large volumes of AI processing.</p><h2>The first requests may be almost invisible</h2><p>Mandiant also saw attackers testing stolen keys by limiting a model&#8217;s response to just one output token, a small unit of text. It reminded me of <a href="https://stripe.com/resources/more/what-is-card-testing-fraud-heres-how-you-can-protect-your-business">credit card testing</a>. Someone steals a card number, runs a tiny charge to see whether it works, then tries to spend more. Here, the attacker is checking whether a stolen AI key works before using it more heavily.</p><p>A spending dashboard might barely move during those first checks. The requests can still show up in API logs, but a team watching only for a sudden jump in costs could miss the testing altogether.</p><p>Credit card companies don&#8217;t just watch for large purchases. They also look for unusual activity, like transactions from unfamiliar locations or sudden changes in spending habits.</p><p>We should be looking for similar warning signs with AI services. A model being used by an application that never called it before, a new API key, unexpected permission changes or requests coming from somewhere unfamiliar. Any one of those could be legitimate. But together? I&#8217;d want someone looking into it.</p><p>I covered a related problem in <a href="https://firstorder.zenone.org/p/the-zombie-agent-how-a-rebrand-killed">The Zombie Agent: How a Rebrand Killed Security</a>: exposed model keys, resale and bills that could keep growing through auto-refill. Mandiant&#8217;s report adds evidence from actual investigations.</p><p>I&#8217;d want to know how quickly we can spot those early requests, identify the stolen credential and stop the spending before it gets out of hand.</p><h2>An alert isn&#8217;t the same as a stop</h2><p>The timing of the other announcement caught my attention.</p><p>On October 30, three days before Mandiant&#8217;s report, <a href="https://cloud.google.com/blog/topics/cost-management/google-cloud-spend-caps-now-generally-available/">Google Cloud announced that spending caps were generally available</a>. For eligible services, a cap can pause new usage when estimated costs exceed a set monthly amount.</p><p>That&#8217;s different from a budget email. An email tells someone a limit was crossed. Somebody still has to read it and act.</p><p><a href="https://docs.cloud.google.com/billing/docs/how-to/budgets-spend-caps">Google&#8217;s documentation</a> limits each cap to one eligible service in one project. Eligible services currently include Gemini API, the Gemini Enterprise Agent Platform, Cloud Run and Cloud Run functions. Requests already running can finish and add charges. Storage and other ongoing resource costs also continue. Google&#8217;s announcement says requests using reserved model capacity, called Provisioned Throughput, keep running too.</p><p>Enforcement isn&#8217;t instantaneous, so charges can exceed the cap. At the start of the next monthly budget period, Google automatically lifts the block and the service can accept requests again. If the stolen credential is still valid, the attacker can resume using it.</p><p><a href="https://developers.openai.com/api/docs/guides/spend-limits">OpenAI&#8217;s API documentation</a> makes a similar distinction between an alert and a hard spending limit. With an enforced hard limit, API requests are rejected when tracked spending reaches the applicable monthly amount. There&#8217;s still a delay before enforcement fully takes effect, and setting the limit too low can interrupt legitimate production requests.</p><p>So I wouldn&#8217;t be satisfied with a screenshot showing a budget has been configured. I&#8217;d want to know which service it covers, whether it actually blocks requests and what happens when it blocks work the company needs.</p><p>There&#8217;s another question: who can change the cap? If someone with compromised cloud privileges can enable expensive services, create fresh credentials or alter the controls, the spending limit needs an owner and an audit trail too.</p><h2>What I&#8217;d test before the next bill arrives</h2><p>I&#8217;d give security, engineering and finance one exercise. No elaborate red-team campaign required.</p><p>Suppose a valid service-account credential starts calling an expensive AI model in a project that normally uses a cheaper one. The requests are coming from somewhere you don&#8217;t recognize. Spending begins climbing.</p><p>I&#8217;d ask the teams to show me the actual path from first signal to stopping the use:</p><ol><li><p><strong>Who sees it first?</strong> Is there a warning for a new model service, a new key, unexpected traffic or a sharp cost increase? Does it reach somebody who can investigate, rather than sit in a billing inbox until Monday?</p></li><li><p><strong>What identifies the account?</strong> Can responders connect the model requests and charges to the project, credential and user or service account responsible? Can they distinguish legitimate application traffic from requests the organization didn&#8217;t authorize? Are the necessary logs enabled? Mandiant notes that the Data Access logs needed to record model requests on Google&#8217;s Gemini Enterprise Agent Platform are off by default.</p></li><li><p><strong>Who can stop it?</strong> Name the person or on-call role authorized to revoke the key, disable an account, pause a service or enforce a provider-supported spending limit. The right action depends on which credential was stolen and what other access the attacker created.</p></li><li><p><strong>What breaks when it stops?</strong> A shared key might also serve a customer-facing application. Have a replacement credential or clean project ready where practical. Decide who approves restoration after the team has removed the attacker&#8217;s access, including any additional accounts or keys the attacker created.</p></li></ol><p>I&#8217;d also want the team to preserve key-creation events, model-request logs, records showing who enabled the service and billing data. Collect what&#8217;s needed without delaying the work to stop the abuse. Rotating the first stolen key won&#8217;t remove another account the attacker created earlier.</p><p>This is where finance belongs in the incident process. Finance can say what charges appear, which billing account they belong to and when the unusual usage started. Security can investigate the access. Engineering can keep the legitimate application working or explain what stopping it will interrupt.</p><p>The cloud provider may have a process to review disputed usage. Ask about it promptly, keep the supporting records and confirm what the invoice will actually require. The Mandiant figures alone don&#8217;t tell us whether any particular victim recovered the charges.</p><p>Mandiant&#8217;s cases don&#8217;t establish that every organization is facing a six-figure AI bill. They establish that criminals have found a way to turn stolen model access into charges quickly enough that the response time matters.</p><p>I&#8217;d rather test that handoff now than find out who owns it when finance asks why yesterday&#8217;s model usage cost more than the month&#8217;s budget.</p><div><hr></div><h3>Resources</h3><ul><li><p><a href="https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models/">Google Threat Intelligence Group and Mandiant, </a><em><a href="https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models/">The Rise of LLMJacking</a></em><a href="https://cloud.google.com/blog/topics/threat-intelligence/llmjacking-monetizing-compromised-cloud-models/">, November 2, 2026</a>. Original incident investigations, entry paths, stolen-key resale and reported usage charges.</p></li><li><p><a href="https://cloud.google.com/blog/topics/cost-management/google-cloud-spend-caps-now-generally-available/">Google Cloud, </a><em><a href="https://cloud.google.com/blog/topics/cost-management/google-cloud-spend-caps-now-generally-available/">Spend Caps Now Generally Available</a></em><a href="https://cloud.google.com/blog/topics/cost-management/google-cloud-spend-caps-now-generally-available/">, October 30, 2026</a>. Product announcement describing service-level controls.</p></li><li><p><a href="https://docs.cloud.google.com/billing/docs/how-to/budgets-spend-caps">Google Cloud documentation, </a><em><a href="https://docs.cloud.google.com/billing/docs/how-to/budgets-spend-caps">Manage Spend Cap Budgets</a></em>. Eligible services, enforcement behavior, exclusions and limitations.</p></li><li><p><a href="https://developers.openai.com/api/docs/guides/spend-limits">OpenAI API documentation, </a><em><a href="https://developers.openai.com/api/docs/guides/spend-limits">Spend Limits</a></em>. Differences between spend alerts and enforced hard limits, including enforcement delay and application impact.</p></li></ul>]]></content:encoded></item><item><title><![CDATA[The Agent Went Off-Script. Is That a Security Incident?]]></title><description><![CDATA[An AI agent can cross a boundary without an attacker. Your incident team still needs a way to stop it, reconstruct the action and authorize a restart.]]></description><link>https://firstorder.zenone.org/p/the-agent-went-off-script-is-that</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-agent-went-off-script-is-that</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 27 Oct 2026 15:31:11 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!HRX8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>On October 16, NIST, NSA and CISA issued voluntary guidance for training and evaluating highly capable agents. One concern is an agent taking a significant action nobody instructed it to take and crossing a control boundary.</p></li><li><p>For ordinary business agents, I&#8217;d add a response path for consequential control failures, even before investigators know the cause.</p></li><li><p>Stopping the agent is only containment. Preserve the evidence, investigate the action and name who can approve a restart.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!HRX8!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!HRX8!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!HRX8!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:209391,&quot;alt&quot;:&quot;Illustration of a records panel, a lever switch and a separate gold key.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/222188317?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="Illustration of a records panel, a lever switch and a separate gold key." title="Illustration of a records panel, a lever switch and a separate gold key." srcset="https://substackcdn.com/image/fetch/$s_!HRX8!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!HRX8!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7bd00b2f-4d24-45af-be81-224e070a3855_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">An intact evidence trail and an independent restart decision matter after containment.</figcaption></figure></div><p>Imagine a procurement agent that can compare suppliers, request quotes and prepare purchase orders. Its instructions say that a person must approve any new supplier before an order goes out.</p><p>Now suppose the agent adds a new supplier and places an order anyway. The credentials are legitimate. The purchasing application accepted the request. Nobody has found a stolen password or an attacker-controlled prompt.</p><p>Those facts don&#8217;t tell you whether the purchase was authorized. They tell you the software let it happen.</p><p>Procurement has to work out what the order commits the company to. Engineering needs to find out why the approval step failed. Security has to check whether someone manipulated the agent or bypassed a control.</p><p>All three have a job. I&#8217;d want the company to decide in advance who can declare an incident and stop the agent while those teams work out what happened.</p><h2>What the new federal guidance actually says</h2><p>On October 16, <a href="https://www.nist.gov/news-events/news/2026/10/key-practices-incident-readiness-and-response-super-intelligence-training">NIST, the NSA and CISA published ten voluntary practices</a> for incident readiness and response while training and evaluating what they call <em>super intelligence</em> agents. The document is aimed primarily at organizations running very capable agents on long tasks with limited human supervision. It isn&#8217;t a new compliance requirement for every company deploying an AI assistant.</p><p>The <a href="https://www.nist.gov/document/keypracticesforincidentreadinessandresponseinsuperintelligence-trainingandevaluation">joint document</a> defines a <em>super intelligence (SI) control incident</em>. In plain English, the agent takes a significant action that its operator didn&#8217;t instruct it to take, and the action compromises a boundary set to limit the task or maintain human oversight. The agencies also recognize near-misses, where the risk was reasonably foreseeable or imminent but didn&#8217;t become an incident.</p><p>There&#8217;s an unusual distinction in the definition. An action directed by an adversary, including through prompt injection, isn&#8217;t <em>uninstructed</em> under this specific category. The agencies treat that as a different cause. A prompt-injection attack can still be a security incident. And an uninstructed agent action can also create a traditional cybersecurity incident if it damages a system or exposes information.</p><p>So I wouldn&#8217;t copy the federal label into a corporate runbook and treat it as a universal classification rule. I&#8217;d borrow the part that changes the investigation: <strong>the absence of an attacker doesn&#8217;t make a consequential control failure harmless.</strong></p><p>Traditional incident response already covers violations of security policy, compromised systems and threats to information. The new guidance builds on that foundation. I&#8217;d give <em>agent-control</em> failures a clear path through the existing response process. Whether a particular failure is a cybersecurity incident still depends on the action and the boundary it crossed.</p><h2>Classify the action before arguing about its cause</h2><p>Return to the procurement example. The first report says, &#8216;The agent sent an unapproved order.&#8217; At that point, the team may not know whether someone altered its instructions, the approval check was broken or the agent found another way to use an authorized tool.</p><p>It doesn&#8217;t need to solve that question before deciding whether the agent should keep placing orders.</p><p>I&#8217;d use two questions in the first response:</p><p><strong>What did the agent actually do, and what could it still do?</strong> Capture the transaction, the amount, the recipient and the systems the agent can reach. Check whether the same credentials or workflow can repeat the action. If other agents share the same tool or policy, they may need to be paused too.</p><p><strong>Which boundary was supposed to stop it?</strong> Find the business rule, the approval step and the technical enforcement point. If nobody can identify the rule, the problem may be an authority gap rather than a bypass. I wrote about that earlier in <a href="https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company">Your AI Agent Can Commit the Company</a>. This case starts after the action has already happened.</p><p>The incident commander should own the response process. The business owner should assess what the action committed the company to and whether it can be reversed. Engineering should reconstruct the agent and tool behavior. Security should investigate whether access or controls failed and whether an adversary was involved. Legal and privacy join when the facts call for them.</p><p>Those roles can overlap in a small company. The ownership of the decision can&#8217;t disappear between them.</p><p>I wouldn&#8217;t classify every odd answer or harmless tool retry as an incident. Severity should follow the real or reasonably imminent consequence, the authority crossed and the ability to contain the behavior. A blocked attempt with no exposure may be a near-miss. A completed external payment, irreversible data change or unauthorized disclosure deserves a different response. The classification can change as the investigation develops, and the team should record why.</p><h2>Your chat transcript won&#8217;t be enough</h2><p>The joint guidance spends real attention on evidence. It calls for preserving configurations, environment state, reliable timestamps, tamper-resistant logs, tool calls that succeeded <strong>and</strong> those that were rejected, and enough information to reconstruct an agent&#8217;s activity.</p><p>This matters. If the only record is a final answer saying the task is complete, investigators may never see the call the agent attempted, the approval it skipped or the retry that eventually succeeded.</p><p>For an enterprise agent, I&#8217;d want a reconstructable chain: who assigned the task, the instruction and policy versions, the agent identity, each tool request and response, the approval decision, the resulting external transaction and the time each step occurred. Capture the surrounding application logs too. <a href="https://morphic.zenone.org/p/who-answers-for-the-agent-ai-accountability">A model&#8217;s self-explanation</a> isn&#8217;t a substitute for the tool and system records.</p><p>Preserve that evidence somewhere the affected agent can&#8217;t change it. If the agent used credentials that might still permit harmful actions, restrict or revoke them as the incident warrants. Don&#8217;t let an enthusiastic cleanup script erase the run state the investigators need.</p><p>There&#8217;s a practical constraint here. Some models and hosted services don&#8217;t expose every internal reasoning trace. You can&#8217;t preserve data you never had. That makes independent records of tool calls, decisions and effects even more important. It also makes observability a purchasing and design question, not something the incident team can retrofit after the fact.</p><p>The goal isn&#8217;t to reproduce every token. It&#8217;s to establish what the agent was authorized to do, what it tried to do and what the connected systems actually did.</p><h2>Who gets to turn it back on?</h2><p>Stopping an agent is the easy part to put on a diagram. Recovery is where the organization has to make a harder decision.</p><p>The federal guidance says to verify the repaired controls before restarting. It recommends rebuilding potentially compromised environments from trusted versions, checking that logging and restrictions work, replaying incident actions against updated protections and documenting the remaining risk. It also calls for leadership to be satisfied with the risks before a high-risk training or evaluation run starts or resumes.</p><p>For ordinary business agents, I&#8217;d make the restart decision smaller and more concrete. Before returning the procurement agent to service, the team should be able to show that the unapproved order path now fails. The approval gate should reject the same action when replayed in a controlled test. The agent&#8217;s credentials should permit only the tasks it still needs. Monitoring should catch a repeat attempt. Procurement should confirm how any outstanding order was handled.</p><p>Then name the person accepting the residual business risk. Security can verify the technical controls. Engineering can attest to the fix. Neither should silently decide that an unauthorized financial commitment is acceptable to procurement.</p><p>A green test run isn&#8217;t proof an agent will never make another mistake. It&#8217;s evidence that the specific failure path was addressed, alongside whatever broader testing the severity calls for.</p><p>The distinction also matters if the agent caused damage without breaking a technical permission. A workflow might allow the action because someone granted authority too broadly. In that case, restarting with the same policy and a new prompt is not much of a recovery plan.</p><h2>Run the exercise while nothing is broken</h2><p>Pick one agent with permission to send something outside the company or change a consequential record. Use an isolated test environment with no production access or real external recipients.</p><p>Create a harmless test case in which the agent completes one action outside its approved scope. Capture the actual tool and approval records, then give the response team the event without identifying its cause. Ask someone to assign severity using the information available at that moment. Have engineering reconstruct the action from the records. Have the business owner identify the consequence that action would have had in production. Then ask who can pause the agent, who can reverse the action and who signs off on restarting it.</p><p>If the drill stops at &#8216;we can disable the agent,&#8217; keep going. You haven&#8217;t tested the investigation or the recovery decision.</p><p>Before the exercise ends, have the person approving the restart write down what has to change and what evidence they&#8217;d need before turning the agent back on.</p><div><hr></div><h3>Resources</h3><ul><li><p><a href="https://www.nist.gov/document/keypracticesforincidentreadinessandresponseinsuperintelligence-trainingandevaluation">NIST, NSA and CISA, </a><em><a href="https://www.nist.gov/document/keypracticesforincidentreadinessandresponseinsuperintelligence-trainingandevaluation">Key Practices for Incident Readiness and Response in Super Intelligence Training and Evaluation</a></em><a href="https://www.nist.gov/document/keypracticesforincidentreadinessandresponseinsuperintelligence-trainingandevaluation">, October 16, 2026</a>. Voluntary joint guidance, especially definitions and practices 2, 5, 6 and 7.</p></li><li><p><a href="https://www.nist.gov/news-events/news/2026/10/key-practices-incident-readiness-and-response-super-intelligence-training">NIST announcement of the joint guidance, October 16, 2026</a>. Describes audience, scope and five-stage lifecycle.</p></li><li><p><a href="https://csrc.nist.gov/pubs/sp/800/61/r3/final">NIST SP 800-61 Rev. 3, </a><em><a href="https://csrc.nist.gov/pubs/sp/800/61/r3/final">Incident Response Recommendations and Considerations for Cybersecurity Risk Management</a></em><a href="https://csrc.nist.gov/pubs/sp/800/61/r3/final">, April 2025</a>. The existing incident-response framework on which the joint guidance builds.</p></li></ul><p><em>Source and scope note: The hypothetical procurement incident, the proposed enterprise response roles and the restart checklist are my operational interpretation. They&#8217;re not incidents documented in the federal guidance or obligations imposed on all enterprises.</em></p>]]></content:encoded></item><item><title><![CDATA[Your AI Workflow Is on the Internet. Who's Responsible for It?]]></title><description><![CDATA[Censys found more than 113,000 exposed n8n servers. Leaders need to know which ones belong to them and what those systems can access.]]></description><link>https://firstorder.zenone.org/p/your-ai-workflow-is-on-the-internet</link><guid isPermaLink="false">https://firstorder.zenone.org/p/your-ai-workflow-is-on-the-internet</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 20 Oct 2026 15:31:14 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!JcqL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Censys found 113,926 n8n servers reachable from the public internet. That doesn&#8217;t mean all of them were hacked or even vulnerable.</p></li><li><p>The real concern is what an automation server can access: email, customer records, databases and other systems connected through stored credentials.</p></li><li><p>I&#8217;d want every exposed automation service to have a named business owner, a reason to be public and evidence that its software and connected accounts have been checked.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!JcqL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!JcqL!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!JcqL!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/f8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2015125,&quot;alt&quot;:&quot;A compact computer connected by cable through a wall to a file organizer holding folders and keys.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/221209003?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="A compact computer connected by cable through a wall to a file organizer holding folders and keys." title="A compact computer connected by cable through a wall to a file organizer holding folders and keys." srcset="https://substackcdn.com/image/fetch/$s_!JcqL!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!JcqL!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff8cd562e-750c-47af-ab7c-15dc355dde10_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-GK1x3M" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-GK1x3M" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption"></figcaption></figure></div><p>Censys found 113,926 n8n servers exposed to the public internet. That&#8217;s a lot of automation sitting where anyone on the internet can find the service.</p><p>Before that number becomes a headline about another security disaster, there&#8217;s an important distinction. A server being reachable doesn&#8217;t mean somebody broke into it. Some workflows need public webhooks or forms to do their job. Others may be exposed without anyone realizing it.</p><p>I&#8217;m more interested in a question executives should be able to answer without calling six different teams: <strong>Which of those systems belong to us, and who is responsible for them?</strong></p><p>n8n is a workflow automation tool. It can connect AI models to other applications, but workflows don&#8217;t have to use AI. A workflow might receive a form, update a customer record, create a ticket or send a message. Those steps often require credentials for the connected services.</p><p>So if someone compromises a badly configured automation server, the damage may not stop at the server. It depends on what accounts the workflows can access and what those accounts are allowed to do.</p><p>That&#8217;s a different conversation from asking whether your employees are using AI.</p><h2>What the research actually says</h2><p>On October 14, <a href="https://censys.com/blog/state-of-the-internet-report-2026/">Censys published findings</a> from its 2026 State of the Internet Report. Its measurements showed that the number of internet-facing hosts exposing AI-related tools grew more than 63% between October 2025 and July 2026.</p><p>The most widely exposed product in that group was n8n, with 113,926 hosts.</p><p><strong>Of the n8n hosts whose software versions Censys could identify, </strong>76% reported a version in the affected range of at least one recent critical vulnerability.</p><p>Read that carefully. The 76% doesn&#8217;t cover every n8n host. It covers the subset that reported a readable version number. And a version being in an affected range doesn&#8217;t tell us whether every condition for exploiting a particular bug is present.</p><p>These are worldwide measurements of internet-facing hosts, not counts of distinct companies or confirmed breaches.</p><p>That gives security teams a place to start: find their own exposed services, then check whether those deployments meet the conditions for a known vulnerability.</p><h2>One workflow can reach a surprising amount of the business</h2><p>Consider a hypothetical workflow built by a support team. A customer submits a request. The workflow reads the form, looks up the customer in a database, opens a support ticket and sends an update to a chat channel.</p><p>Perfectly reasonable automation.</p><p>To do that work, the server needs a way to authenticate to those systems. Maybe the customer database connection can only read a few fields. Maybe the support account can create tickets but can&#8217;t delete them. Or maybe somebody used a broad administrator credential because it was faster to get the workflow running.</p><p>Those are very different risks, even if the workflow looks identical to the employee using it.</p><p>If the server is compromised, what could someone do using the connections it already has? Read customer records? Change tickets? Send messages that appear to come from the company? Reach files or internal services?</p><p>Leadership needs to know the answers for its own deployment.</p><p>In <a href="https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company">Your AI Agent Can Commit the Company</a>, I wrote about the business limits that have to exist before an agent is allowed to take action. This is the infrastructure side of that problem. Someone has to know what the server can reach before they can claim those limits are enforced.</p><h2>A version number is only the beginning</h2><p>In a <a href="https://blog.n8n.io/security-advisory-20260108/">January 8, 2026 security advisory</a>, n8n described a file-access vulnerability affecting self-hosted versions 1.65 through 1.120.4. An unauthenticated person could read files from the host if it ran an active, publicly accessible workflow with a Form Submission trigger and a Form Ending node configured to return a binary file. The vendor said the issue was fixed in version 1.121.0, released November 18, 2025.</p><p>The advisory is an example of why configuration matters. Censys&#8217;s public summary doesn&#8217;t identify which vulnerabilities account for its 76%.</p><p>So a review can&#8217;t stop at finding an old version and declaring a breach. Check the exact configuration, the vendor&#8217;s advice, whether the update is installed and whether any sensitive data could have been exposed.</p><p>The same vendor provides a <a href="https://docs.n8n.io/hosting/securing/security-audit/">built-in security audit</a> that can flag things such as unprotected webhooks, outdated versions, risky nodes and unused credentials. That&#8217;s helpful, but you still need to check how the service is exposed from outside and what its connected accounts can do.</p><p>And while n8n is the clearest example in this research, the ownership question applies to other automation tools too. Some run in your cloud account, some are managed by a vendor and some were built by one person to solve a problem everyone else was too busy to address.</p><h2>What I&#8217;d ask before calling this handled</h2><p>I wouldn&#8217;t ask the board to memorize n8n version numbers. However, I would expect the security and business owners to be able to show a few things.</p><p><strong>First, find the services from the outside.</strong> Use an approved external exposure assessment, then compare what it finds against your cloud inventory, vendor list and known automations. If you only look at systems IT says it owns, you may miss the one someone stood up outside the normal process. Confirm ownership before touching or testing anything that&#8217;s not yours.</p><p><strong>Next, name the business owner and the people responsible for running the service.</strong> The business owner should be able to explain why it&#8217;s needed and what it&#8217;s allowed to access. Name who handles patching and expired credentials, who responds to an incident and who takes over if the person who built it leaves.</p><p><strong>Then, explain why it has to be reachable.</strong> Some workflows legitimately accept public requests. That doesn&#8217;t mean the editor, administration screens and other functions should automatically be accessible to everyone. Check which parts require public access and which can be restricted.</p><p><strong>Finally, inspect what it can reach and verify the fix.</strong> Inventory the credentials and the permissions behind them. Remove access the workflow doesn&#8217;t need. Review the vendor&#8217;s security advisories, upgrade or apply the supported mitigation, then check again from outside. If there&#8217;s reason to suspect a compromise, investigate it before assuming a patch by itself closes the incident.</p><h2>The part the inventory doesn&#8217;t tell you</h2><p>I wrote about supplier visibility in <a href="https://firstorder.zenone.org/p/you-cant-patch-a-supply-chain-you">You Can&#8217;t Patch a Supply Chain You Can&#8217;t Inventory</a>. You need to know what&#8217;s installed before you can deal with it.</p><p>Here, there&#8217;s another question after you find the system: what authority did you give it?</p><p>Two automation servers can look equally exposed from the internet. One has limited credentials and accepts only a specific public request. The other can read a customer database and change records across several business applications.</p><p>I wouldn&#8217;t treat those as the same risk, and I wouldn&#8217;t be satisfied with a report that simply counts both as exposed.</p><p>Censys has given us a reason to look. The job for leadership is to find out what&#8217;s actually running, who owns it and how much access sits behind it.</p><p>If I were reviewing the results with a CISO, that&#8217;s the information I&#8217;d want on the first page. Not a green box that says the AI tools are approved.</p><h2>Resources</h2><ul><li><p><a href="https://censys.com/blog/state-of-the-internet-report-2026/">Censys, 5 Key Takeaways From the 2026 State of the Internet Report, October 14, 2026</a></p></li><li><p><a href="https://censys.com/2026-state-of-the-internet-report/">Censys, 2026 State of the Internet Report overview</a></p></li><li><p><a href="https://blog.n8n.io/security-advisory-20260108/">n8n, January 8, 2026 security advisory</a></p></li><li><p><a href="https://docs.n8n.io/hosting/securing/security-audit/">n8n documentation, Security audit</a></p></li></ul>]]></content:encoded></item><item><title><![CDATA[Microsoft Defended the Numbers. Who Defends the Decisions?]]></title><description><![CDATA[Microsoft's visa response doesn't settle the questions about its layoffs, hiring or the expertise it lost.]]></description><link>https://firstorder.zenone.org/p/microsoft-defended-the-numbers-who</link><guid isPermaLink="false">https://firstorder.zenone.org/p/microsoft-defended-the-numbers-who</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 13 Oct 2026 15:30:46 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!mW-Z!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>I agree with JD Vance and the White House that immigration programs shouldn&#8217;t be used to displace qualified American workers.</p></li><li><p>Microsoft&#8217;s H-1B filing statistics challenge a simple comparison with layoffs. They don&#8217;t establish whether particular jobs were eliminated and later filled with comparable work.</p></li><li><p>The larger question for Microsoft&#8217;s leaders is what their workforce decisions achieved, including the expertise lost through layoffs and office mandates.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!mW-Z!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!mW-Z!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!mW-Z!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/f4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2153722,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/220060390?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!mW-Z!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!mW-Z!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff4756db4-fb8b-4870-959b-b8adb0a35c64_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-GK1x3M" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-GK1x3M" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">An empty desk tells you someone left. It doesn't tell you whether their work disappeared, or what the company lost when they did.</figcaption></figure></div><p>On October 8, Vice President JD Vance <a href="https://rollcall.com/factbase/trump/transcript/donald-trump-press-conference-jd-vance-visa-fraud-october-8-2026/">accused Microsoft</a> of laying off American workers while sponsoring foreign workers through immigration programs. The administration announced it was suspending Microsoft from PERM, the federal labor-certification process used for certain employment-based green cards, amid an ongoing investigation.</p><p>I agree with Vance and the White House on the concern that matters. <strong>Companies shouldn&#8217;t be allowed to lay off qualified American workers and use immigration programs to replace them with cheaper, more dependent labor.</strong> If that happened at Microsoft, it deserves a serious investigation and appropriate consequences.</p><p>My concern isn&#8217;t with people working here on visas. Many are skilled employees whose immigration status depends heavily on an employer. They didn&#8217;t make Microsoft&#8217;s hiring or layoff decisions. Those decisions belong to its leadership.</p><p>Microsoft <a href="https://blogs.microsoft.com/on-the-issues/2026/10/08/microsoft-comment-on-recent-immigration-questions/">responded</a> that about 80% of its roughly 6,000 H-1B filings last fiscal year involved existing employees. It said the remaining filings concerned new employees already legally in the United States, a number equivalent to about 1% of its U.S. workforce. Microsoft also said it pays H-1B workers comparably to employees doing similar work.</p><p>Fine. Those details matter. But they don&#8217;t answer the question I&#8217;m interested in.</p><p><strong>Did Microsoft eliminate jobs held by qualified American workers and then hire others to perform substantially the same work?</strong></p><h2>Microsoft&#8217;s numbers don&#8217;t answer that question</h2><p>Let&#8217;s separate the numbers first. An H-1B petition can involve an existing employee. A PERM labor certification concerns a different process: an employer&#8217;s sponsorship of a worker for permanent residency. Neither figure, by itself, tells us which jobs were eliminated or who took over the work.</p><p>That distinction matters because a misleading comparison makes a legitimate concern easier to dismiss. But Microsoft&#8217;s response stops short of the employment records needed to examine the allegation.</p><p>A new Microsoft hire who already lives in the United States is still a new Microsoft hire. Their location tells us nothing about whether the job resembles one the company recently eliminated. And saying that two employees doing comparable work earn comparable wages doesn&#8217;t establish whether a qualified U.S. worker was available or considered for a particular opening.</p><p>I&#8217;d want investigators to match the positions Microsoft eliminated against positions it later filled. Not just job titles, which companies can change. The actual duties, location, timing, required skills, recruiting records and compensation matter. So does whether the underlying work continued on another team.</p><p>The rules also need to be applied to the right program. For a <a href="https://www.dol.gov/agencies/eta/foreign-labor/programs/permanent">PERM labor certification</a>, the Department of Labor generally requires an employer to show that there aren&#8217;t sufficient qualified and available U.S. workers for the position and that hiring the foreign worker won&#8217;t adversely affect comparable U.S. workers&#8217; wages and conditions. An H-1B petition isn&#8217;t the same application, and Microsoft&#8217;s H-1B percentages don&#8217;t establish compliance with every PERM requirement.</p><p>We don&#8217;t know what nonpublic evidence the administration has. The announced suspension and investigation aren&#8217;t a finding that Microsoft broke the law. They&#8217;re also not something Microsoft&#8217;s short statement can resolve on its own.</p><p><strong>The records could clear Microsoft. They could also substantiate the allegation.</strong> Preserve the original hiring, layoff and immigration records and let investigators test them. If improper displacement is established, the people who made or approved those decisions should be held accountable.</p><h2>If the jobs disappeared, why didn&#8217;t the work?</h2><p>Some layoffs are necessary. Demand changes, products fail and companies can&#8217;t keep every role indefinitely. What bothers me is when executives point to a smaller payroll as though that alone proves they made the right decision.</p><p>Consider a company that eliminates a team responsible for keeping critical systems running. The salaries disappear from the budget, but the systems still need maintaining. Someone else inherits the alerts, the on-call shifts and the problems those employees used to prevent.</p><p>The company cut the people. Did it actually eliminate the work?</p><p>That&#8217;s a hypothetical, not a claim about a particular Microsoft team. It&#8217;s also something leadership can measure. Which responsibilities were transferred? Did customers wait longer? Did the remaining employees take on more risk? Was the expertise later replaced at a higher cost?</p><p>I&#8217;ve written about the operational cost of losing experienced people in <a href="https://firstorder.zenone.org/p/the-security-team-you-cut-doesnt">The Security Team You Cut Doesn&#8217;t Come Back on Incident Day</a>. The point applies beyond security. A cut can look successful for months before the cost becomes visible.</p><p>I would expect Microsoft&#8217;s board to examine that cost, not just the immediate savings. Otherwise, we can&#8217;t tell whether a layoff improved the business or simply moved the work and its risks onto fewer people.</p><h2>Return to office: What happens when experienced people leave?</h2><p>There&#8217;s another Microsoft decision that deserves the same scrutiny.</p><p>In <a href="https://blogs.microsoft.com/blog/2025/09/09/flexible-work-update/">September 2025</a>, Microsoft announced that employees living within 50 miles of an office in the Puget Sound area would generally be expected on site three days a week by the end of February 2026. Exceptions were possible, and business groups could set different expectations. The company said the change was about collaboration and results, not reducing headcount.</p><p>That may be the goal. I&#8217;d still want to see who left and whether the collaboration improved.</p><p><a href="https://bfi.uchicago.edu/insights/return-to-office-and-the-tenure-distribution/">Research on earlier return-to-office mandates</a> at Microsoft, Apple and SpaceX found that longer-tenured and senior employees were more likely to leave after those policies. A <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5031481">separate study of more than three million technology and finance workers</a> found higher turnover following office mandates, particularly among senior and skilled employees, along with longer times to fill vacancies.</p><p>Neither study establishes why Microsoft adopted its September 2025 policy or what happened after that particular rollout. In-person work can help some teams. But the studies give executives a real risk to measure, especially after major layoffs: experienced employees may leave voluntarily, taking knowledge the company never intended to lose.</p><p>I&#8217;d want Microsoft to report retention by tenure and role, the cost of replacing departing employees and evidence that teams are actually working better together. Badge swipes tell you who entered a building. They don&#8217;t tell you whether the work improved.</p><p>If the policy delivers better results, show them. If it doesn&#8217;t, change it.</p><h2>Following the crowd isn&#8217;t leadership</h2><p>There&#8217;s a broader management problem here. When a large technology company announces layoffs, it&#8217;s easy for other executives to point to the news as proof that their own cuts make sense.</p><p>Maybe their business really does need to shrink. But somebody else&#8217;s decision isn&#8217;t the analysis. Leaders still need to show what work disappeared, what alternatives they tried and why the reduction was the right choice for their company.</p><p>Other approaches exist. In <a href="https://www.fortune.com/2026/05/28/costco-ceo-ron-vachris-artifical-intelligence-not-stealing-jobs-tech-improving-work-retail-giant-success/">Fortune&#8217;s May 2026 reporting</a>, Costco CEO Ron Vachris said the company had moved employees into other responsibilities as technology changed their tasks, rather than displacing them. He also explained that Costco&#8217;s business was growing faster than those changes eliminated work. That distinction matters. A growing retailer and a software company facing different pressures aren&#8217;t directly comparable.</p><p>Still, it demonstrates that executives have more than one choice. They can retrain people, move them to work that&#8217;s growing or redesign a job before eliminating it. Sometimes those options won&#8217;t be enough.</p><p>I&#8217;d respect a leadership team that could show why it rejected them. <strong>If your justification for layoffs is that another big company did it, you haven&#8217;t made your case.</strong></p><h2>Show us what happened</h2><p>Microsoft has supplied a specific explanation of its H-1B filing totals. The White House has made a serious allegation and announced an investigation. Neither a press conference nor a corporate statement supplies the job-by-job evidence needed to settle the dispute.</p><p>So I&#8217;d like Microsoft&#8217;s leadership to answer a more useful set of questions. Which jobs actually disappeared? Which duties remained? Were comparable positions filled afterward, and how were qualified U.S. workers considered? What did the company save after accounting for the expertise it lost?</p><p>Those are fair questions for any large employer.</p><p><strong>If the records clear Microsoft, the findings should say so. If they establish improper displacement, the response should reach the people responsible for it.</strong> Either way, I&#8217;d rather see the evidence than another round of carefully chosen statistics.</p><h2>Resources</h2><ul><li><p><a href="https://rollcall.com/factbase/trump/transcript/donald-trump-press-conference-jd-vance-visa-fraud-october-8-2026/">JD Vance&#8217;s October 8, 2026 press conference, transcript via Roll Call</a></p></li><li><p><a href="https://blogs.microsoft.com/on-the-issues/2026/10/08/microsoft-comment-on-recent-immigration-questions/">Microsoft&#8217;s October 8, 2026 response on immigration filings</a></p></li><li><p><a href="https://www.dol.gov/agencies/eta/foreign-labor/programs/permanent">U.S. Department of Labor, Permanent Labor Certification</a></p></li><li><p><a href="https://blogs.microsoft.com/blog/2025/09/09/flexible-work-update/">Microsoft&#8217;s September 9, 2025 flexible-work announcement</a></p></li><li><p><a href="https://bfi.uchicago.edu/insights/return-to-office-and-the-tenure-distribution/">University of Chicago Becker Friedman Institute, Return to Office and the Tenure Distribution</a></p></li><li><p><a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5031481">Ding et al., Return to Office Mandates, Brain Drain and Gender Difference</a></p></li><li><p><a href="https://www.fortune.com/2026/05/28/costco-ceo-ron-vachris-artifical-intelligence-not-stealing-jobs-tech-improving-work-retail-giant-success/">Fortune, Ron Vachris on Costco, technology and employment, May 28, 2026</a></p></li></ul>]]></content:encoded></item><item><title><![CDATA[Your AI Pilot Passed. Against What?]]></title><description><![CDATA[Before the pilot starts, agree on what would justify deployment, what would stop it and who can change the test.]]></description><link>https://firstorder.zenone.org/p/your-ai-pilot-passed-against-what</link><guid isPermaLink="false">https://firstorder.zenone.org/p/your-ai-pilot-passed-against-what</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 06 Oct 2026 15:31:29 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!_cQr!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Before an AI pilot starts, write down what work the tool may do, what improvement would justify using it and which failures would stop deployment.</p></li><li><p>Measure the work after people check and correct the output. A fast first draft can still leave the team with more work.</p></li><li><p>If the test changes, keep the original result. Run another test under the new conditions before calling the pilot a success.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!_cQr!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!_cQr!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!_cQr!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2519042,&quot;alt&quot;:&quot;A hand paints a red target circle around an arrow already embedded in a wooden board.&quot;,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/218971908?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="A hand paints a red target circle around an arrow already embedded in a wooden board." title="A hand paints a red target circle around an arrow already embedded in a wooden board." srcset="https://substackcdn.com/image/fetch/$s_!_cQr!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!_cQr!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3d9d2f20-8c20-4d42-a0f1-add7aef23e33_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-GK1x3M" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-GK1x3M" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">Painting a target around the result doesn't prove the test was passed.</figcaption></figure></div><p>If you can&#8217;t show me what would have made your AI pilot fail, I don&#8217;t know what &#8220;passed&#8221; means.</p><p>A demo can show something useful. People can like the tool. The team can learn enough to justify another experiment. None of that tells the executive approving deployment what the tool has actually earned permission to do.</p><p>The trouble starts when we treat those different decisions as one. Someone asks whether the technology looks promising. A few weeks later, the answer has become permission to put it into a workflow people depend on. The pilot never tested that decision.</p><p>On August 7, 2026, the National Institute of Standards and Technology, NIST, announced a <a href="https://www.nist.gov/artificial-intelligence/ai-research/tevv-athlon-framework-evaluating-ai-systems">draft framework for evaluating AI systems</a> and invited comments through October 6, 2026. It starts evaluation with the organization&#8217;s goals and the people who need the results. That&#8217;s a useful place to start. The harder management job is agreeing on what result would justify deployment before the team knows whether the tool can produce it.</p><h2>The demo can choose the test for you</h2><p>Consider a hypothetical pilot for an AI assistant that drafts incident handoff summaries. A responder provides the incident records. The assistant produces a summary for the next person taking over.</p><p>The demo looks good because the output is readable and arrives quickly. Someone suggests measuring how long the assistant takes to produce it. That&#8217;s easy to count, so it becomes the pilot&#8217;s success measure.</p><p>But the receiving responder needs an accurate handoff. They need to know which systems were affected and what actions were actually taken. If someone has to reopen the records, correct invented details and rewrite the summary, the assistant&#8217;s speed tells us very little about the time saved.</p><p>I covered that measurement problem in <a href="https://firstorder.zenone.org/p/the-51-point-gap-your-people-got">The 51-Point Gap: Your People Got Faster. The Work Didn&#8217;t.</a>. Here, there&#8217;s an earlier decision to make. Who gets to define &#8220;done&#8221; before the pilot produces an attractive number?</p><p>It shouldn&#8217;t be decided solely by the person running the demo or the executive who wants the purchase to work out. The person responsible for the workflow has to help define the result. Otherwise, the team can pass a test that the people receiving its work would never have chosen.</p><h2>Write the deployment decision before the pilot</h2><p>In <a href="https://firstorder.zenone.org/p/the-problem-statement-problem">The Problem Statement Problem</a>, I wrote about executing against a brief whose assumptions nobody challenged. An AI pilot needs that same scrutiny, followed by something more specific: written conditions for the decision it will support.</p><p>For the handoff example, I&#8217;d start with a narrow proposal: use the assistant to draft summaries that a responder checks before sharing. The assistant doesn&#8217;t send the summary or decide which response action to take.</p><p>Then I&#8217;d ask that person to agree to a short decision record with the sponsor. It could look like this:</p><ul><li><p><strong>What would we authorize after a pass?</strong> Drafting handoff summaries in the named workflow, with a responder checking each one.</p></li><li><p><strong>What improvement would justify using it?</strong> A stated reduction in time to a checked, usable handoff compared with the current process, without sacrificing accuracy. Count checking and corrections. Record both elapsed time and the time people spend doing the work.</p></li><li><p><strong>Which failures would stop deployment?</strong> Name them: claiming a responder isolated a system when they didn&#8217;t, identifying the wrong affected system or disclosing information to someone who shouldn&#8217;t receive it. Specify what would block deployment even if the summary were later corrected.</p></li><li><p><strong>What will the test include?</strong> Cases chosen separately from the demo, including incomplete records and conflicting information. Use the data access planned for deployment.</p></li><li><p><strong>Who makes the decision?</strong> Name the person responsible for the workflow and the sponsor. Record who can approve a change to the test or accept a remaining risk.</p></li></ul><p>There isn&#8217;t a percentage in this example because I don&#8217;t know the team&#8217;s workload or the consequences of a bad handoff. Those facts should determine the threshold. A round number chosen without that context wouldn&#8217;t tell the executive whether the tool was ready to use.</p><p>Write down the comparison process too. Have experienced responders agree on what each test case&#8217;s records establish and what remains uncertain. Judge both the current process and the AI-assisted process against that standard. Where practical, hide how each summary was produced from the responder judging it. If today&#8217;s handoffs are incomplete, beating them on speed alone wouldn&#8217;t establish that the new process produces usable handoffs.</p><p>This is my recommendation for running the pilot. NIST&#8217;s draft doesn&#8217;t prescribe this decision record or require a particular approval process.</p><h2>Give the test a real chance to fail</h2><p>A test built from the cases that made the demo look good can confirm very little. Include the conditions the tool will encounter when people are busy and the records are imperfect. Keep some cases out of development so the team has to show that its changes work on more than the examples it used to make them.</p><p>NIST&#8217;s September <a href="https://www.nist.gov/publications/aria-evaluation-planning-manual-elements-aria-style-ai-evaluations">ARIA Evaluation Planning Manual</a> describes combining three kinds of testing. Model testing uses prepared inputs to check the system&#8217;s responses. Red teaming asks people to try to make it behave badly. User testing examines how it performs when people use it for its intended task.</p><p>For our handoff pilot, test whether the responder spots a wrong system name and checks a confident statement against the record. Measure how much time that checking takes. A requirement for human review only helps if the review can catch the errors that matter.</p><p>Keep the individual failures visible. A high overall score can hide the one kind of error that makes the workflow unacceptable. In <a href="https://morphic.zenone.org/p/evaluating-agents-you-cant-trust">Evaluating Agents You Can&#8217;t Trust Yet</a>, I examined different tests for different agent roles and looked beyond the aggregate score. That article also described deploying after an automated HOLD. Here, I&#8217;d make the record explicit: approving an exception doesn&#8217;t change the failed test result.</p><p>A small test that finds no serious errors also doesn&#8217;t prove that serious errors won&#8217;t occur. Record how many cases you tested and what kinds of cases they were. If a consequential failure would be rare, the owner needs to know that the pilot may not have had much opportunity to encounter it.</p><p>That can justify more testing or a smaller initial deployment. It can&#8217;t justify quietly replacing &#8220;we didn&#8217;t see it&#8221; with &#8220;it doesn&#8217;t happen.&#8221;</p><h2>Keep the original result when you change the test</h2><p>Early experiments should help a team learn. They may reveal that the proposed workflow is wrong, the comparison is weak or the tool needs a different task. Change the plan when the evidence warrants it.</p><p>But keep a record of what changed and why. An experiment used to develop the tool serves a different purpose from a test used to approve it.</p><p>Suppose the assistant fails on incidents with conflicting records. The team decides it will only draft summaries when the records agree. That might be a sensible restriction. It also changes what the tool would be allowed to do.</p><p>The report should retain the first result, describe the narrower use and show the result of a new test. Include how often the restriction would exclude real incidents and who would handle those handoffs. Count that work when estimating the benefit. Otherwise, the executive may approve a tool expecting help across the workload when the evidence supports only a small part of it.</p><p>The same applies if the team changes the model, its instructions or the review process. Those changes may fix the problem. Test that claim on fresh cases rather than treating the corrected development examples as proof.</p><p>If the sponsor wants to proceed despite a failed condition, record the exception. Name who had authority to accept it, what limits apply and what would end the exception. Calling that a pass would erase information the next owner needs.</p><h2>A pass should name what it authorizes</h2><p>In March, <a href="https://www.nist.gov/news-events/news/2026/03/caisi-signs-mou-gsa-boost-ai-evaluation-science-federal-procurement-through">NIST and the General Services Administration announced work on AI evaluation for federal procurement</a>, including evaluation before deployment and monitoring afterward. That work gives government buyers another reason to ask what a test establishes before a purchase. For our pilot, monitoring during deployment should help the owner catch problems the test missed. It doesn&#8217;t supply evidence missing from the current decision.</p><p>A successful handoff pilot could support using the assistant in the tested workflow with the tested review process. It wouldn&#8217;t establish that the assistant can write customer notifications, operate with broader access or send summaries without review. Each change creates a decision the original pilot may not have tested.</p><p>Before approving your next AI pilot for deployment, ask for the written conditions and the results against them. Read the failures and any changes to the test. The person responsible for the workflow should be able to explain what the evidence supports without relying on the enthusiasm of the project team.</p><p>If the conditions were never written, the pilot may still have taught you something useful. Record what you learned, then agree on the test needed to make the deployment decision.</p><h3>Resources</h3><ul><li><p><a href="https://www.nist.gov/artificial-intelligence/ai-research/tevv-athlon-framework-evaluating-ai-systems">NIST, The TEVV-Athlon Framework for Evaluating AI Systems</a>. Initial public draft announced August 7, 2026; comment deadline October 6, 2026. <a href="https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.200-2.ipd.pdf">Full draft, NIST AI 200-2</a>. The framework starts with organizational objectives and develops an evaluation suited to the intended use.</p></li><li><p><a href="https://www.nist.gov/publications/aria-evaluation-planning-manual-elements-aria-style-ai-evaluations">NIST, ARIA Evaluation Planning Manual: Elements of ARIA-style AI Evaluations</a>, September 18, 2026. Describes evaluations combining model testing, red teaming and user testing.</p></li><li><p><a href="https://www.nist.gov/news-events/news/2026/03/caisi-signs-mou-gsa-boost-ai-evaluation-science-federal-procurement-through">NIST, CAISI signs MOU with GSA to boost AI evaluation science for federal procurement</a>, March 18, 2026. Announces work on evaluations before deployment and monitoring after deployment.</p></li></ul>]]></content:encoded></item><item><title><![CDATA[The Delegation Policy You Never Wrote]]></title><description><![CDATA[Your team is already deciding which thinking goes to AI. If you wait for a policy, the habit may already be the policy.]]></description><link>https://firstorder.zenone.org/p/the-delegation-policy-you-never-wrote</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-delegation-policy-you-never-wrote</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 29 Sep 2026 15:31:31 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!AUqs!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>AI delegation is already happening inside ordinary work, often before leaders decide where the boundary should be.</p></li><li><p>Advanced AI users in Microsoft&#8217;s survey are more likely to pause and decide what belongs with AI versus a human, while repeatable handoff rules remain uncommon.</p></li><li><p>Pick one recurring workflow and find the steps you deliberately handed to AI versus the ones it took over by habit.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!AUqs!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!AUqs!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!AUqs!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png" width="1456" height="813" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/a4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:813,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:8949348,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/217856577?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!AUqs!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!AUqs!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa4f40f25-f5bf-48b7-95cf-a057931e030a_2752x1536.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A stone walkway and a worn dirt path run through a grassy field, showing how repeated use can establish a default route.</figcaption></figure></div><p>An employee asks AI to summarize a report. Then they ask what matters most. A week later they ask which options the team should consider, and eventually the AI recommendation becomes the first version everyone reacts to. No single step feels like a governance decision. Nobody announces that part of the company&#8217;s judgment is moving to a model. The boundary just moves.</p><p>A 2026 article in <em>Organizational Dynamics</em> calls this the formation of AI delegation norms. The author argues that organizations often don&#8217;t design in advance which reasoning tasks people should keep and which they should hand to AI. Employees make useful local choices, leaders model behavior and coworkers copy what seems to work. A pattern becomes normal before formal governance catches up.</p><p>The paper is conceptual. It doesn&#8217;t measure how often this happens across companies, and the author reports that no data were used for the research described. I wouldn&#8217;t use it as proof that every organization has this problem. I think it names a problem that current workplace data makes hard to ignore, and it&#8217;s exactly what I&#8217;m seeing when I&#8217;m brought in to consult with companies.</p><h2>The Policy Forms Before the Policy</h2><p>Microsoft&#8217;s 2026 Work Trend Index combines a survey of 20,000 workplace AI users with separate Microsoft 365 Copilot usage analysis. In that usage analysis, 49% of classified Copilot conversations supported what Microsoft calls cognitive work. Microsoft included analysis and problem-solving in that category, along with evaluation and creative thinking. People are already using AI to analyze and evaluate, not just draft.</p><p>Microsoft found another difference. Fifty-three percent of the survey&#8217;s most advanced AI users, a group Microsoft calls Frontier Professionals, said they pause before starting work to decide what belongs with AI versus a human. Among other AI users, 33% said the same. What matters is the pause. These users stop and decide where the line goes.</p><p>Documented handoffs are much less common. Even among those advanced users, 26% said agent workflows, human handoffs and quality standards were documented and repeatable at team level, 29% at function level and 25% at organization level. The numbers were lower for everyone else.</p><p>BCG found another sign that AI is changing what people do at work. In its 2026 survey of 11,749 workers across 14 markets, 72% said AI had considerably changed the skills expected in their roles. Forty-seven percent said they were spending more time managing and directing AI than doing the underlying work themselves.</p><p>Those numbers don&#8217;t tell us where the human-AI boundary should be. They show why leaders shouldn&#8217;t assume the boundary is waiting for them to define it.</p><h2>Repetition Teaches the Organization</h2><p>Technology doesn&#8217;t simply perform a function. Once it becomes part of daily life, it can also change behavior and what people come to treat as normal. You can see that inside a company.</p><p>A manager starts using AI to prepare a strategy memo, and the team sees the result and copies the practice. A weekly report that used to begin with someone reading the underlying material now begins with an AI summary. A vendor review that used to start with people comparing evidence now starts with an AI-generated recommendation.</p><p>Those may be good changes. But repeated use can settle the question before leadership ever asks it. After six months, asking whether the workflow should begin with an AI recommendation feels like adding friction. The habit has become part of the job.</p><p>That&#8217;s why I wouldn&#8217;t reduce this to an AI policy problem. Policies are documents. Norms are what people actually do when nobody is discussing the policy.</p><h2>This Is a Different Delegation Problem</h2><p>I wrote about the human version of delegation in <a href="https://firstorder.zenone.org/p/the-delegation-illusion">The Delegation Illusion</a>. That problem starts after a leader decides to delegate. The failure is handing over a task without enough of the reasoning and decision criteria for the other person to act well. This problem starts earlier: who decided the reasoning should be delegated at all?</p><p>I wrote about another boundary in <a href="https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company">Your AI Agent Can Commit the Company</a>. That article asks who decides what an agent may actually do: issue a refund, offer a discount, move money or make a promise to a customer.</p><p>An AI system doesn&#8217;t need that kind of authority to change how a company thinks. If the model decides which evidence gets summarized and which recommendation everyone sees first, it can shape the decision long before it has permission to execute anything. That doesn&#8217;t make the workflow wrong. It makes the boundary worth choosing on purpose.</p><h2>Run a Delegation Audit</h2><p>Shabad proposes a Delegation Audit. Here&#8217;s how I&#8217;d apply that idea: pick one recurring workflow where people already use AI. A vendor selection works. So does a hiring review or quarterly planning process.</p><p>Write down the actual steps. For example:</p><ol><li><p>Gather the information.</p></li><li><p>Decide what is relevant.</p></li><li><p>Generate possible options.</p></li><li><p>Compare the options.</p></li><li><p>Recommend a choice.</p></li><li><p>Make the decision.</p></li><li><p>Take the action.</p></li><li><p>Review what happened.</p></li></ol><p>For each step, write down how the work is split between people and AI, and who is accountable for the result. Then ask one question: <strong>Who decided that boundary?</strong></p><p>There are good reasons to let AI do a lot of this work. It can search and compare faster than a person starting from a blank page. This isn&#8217;t an argument for keeping humans in every box. It&#8217;s a way to separate an intentional handoff from a habit.</p><p>For any handoff to AI nobody remembers choosing, ask what the person sees before accepting the output, what should send the work back to a human and when the team last checked whether the split still makes sense. You don&#8217;t need a new committee for this. You need a visible decision about where the thinking happens.</p><h2>Decide Before Habit Decides for You</h2><p>The <em>Organizational Dynamics</em> paper argues that early delegation norms are cheaper to shape than settled ones. That seems reasonable, but the article doesn&#8217;t provide organization-wide experimental evidence for the size of that cost difference. We don&#8217;t need to pretend it does. We know something simpler: once a habit is established, people start expecting the workflow to work that way.</p><p>If a team has spent six months letting AI produce the first recommendation, asking people to return to primary material may feel like extra work. If managers routinely use AI summaries before meetings, reading the original documents may start to look inefficient. If employees are measured on speed while leaders say they should preserve human judgment, speed has a very good chance of winning. That&#8217;s how an unspoken delegation choice becomes part of the job.</p><p>Pick one important workflow this week. Map where the thinking happens now, then keep the delegation that makes sense and change the part that doesn&#8217;t. But make somebody say the boundary out loud.</p><p>If nobody ever decided where the line should be, you don&#8217;t have an AI delegation policy yet. You have a habit.</p><div><hr></div><h3>Resources</h3><ul><li><p>Vsevolod Shabad, <a href="https://www.sciencedirect.com/science/article/pii/S0090261626000392">&#8220;The norm architects: How AI delegation norms form before anyone designs them&#8221;</a>, <em>Organizational Dynamics</em>, 2026. Conceptual article on how AI delegation norms can emerge through repeated organizational behavior; the article reports that no data were used for the research described.</p></li><li><p>Microsoft, <a href="https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization">&#8220;2026 Work Trend Index: Agents, human agency, and the opportunity for every organization&#8221;</a>, May 5, 2026. Survey of 20,000 workplace AI users across 10 markets plus separate Microsoft 365 Copilot usage analysis.</p></li><li><p>Boston Consulting Group, <a href="https://www.bcg.com/press/3june2026-ai-reshaping-jobs-faster-than-companies-reshaping-work">&#8220;AI Is Reshaping Jobs Faster Than Companies Are Reshaping Work&#8221;</a>, June 3, 2026. Global survey of 11,749 workers across 14 markets.</p></li><li><p>The Consilience Project, <a href="https://consilienceproject.org/technology-is-not-values-neutral-ending-the-reign-of-nihilistic-design-2/">&#8220;Technology is Not Values Neutral: Ending the Reign of Nihilistic Design&#8221;</a>, June 26, 2022. Broader argument that technologies shape behavior, attention and values as they become embedded in everyday life.</p></li></ul><p><em>Evidence note: Shabad&#8217;s article supplies the norm-formation framework, not a prevalence estimate. Microsoft&#8217;s survey measures self-reported workplace behavior and organizational conditions; the 49% cognitive-work figure comes from separate Copilot usage analysis. BCG&#8217;s results are self-reported survey findings.</em></p><p><em>Analytical note: Shabad supplies the norm-formation argument and proposes a Delegation Audit. The workflow checklist here applies that idea to the examples above; the recommendations are my synthesis of the sources.</em></p>]]></content:encoded></item><item><title><![CDATA[The 24-Hour Clock Starts Before You Know the Whole Story]]></title><description><![CDATA[The EU now requires manufacturers to warn regulators within 24 hours of becoming aware of an actively exploited vulnerability or severe product-security incident.]]></description><link>https://firstorder.zenone.org/p/the-24-hour-clock-starts-before-you</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-24-hour-clock-starts-before-you</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 22 Sep 2026 15:31:58 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!eY4X!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!eY4X!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!eY4X!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 424w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 848w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 1272w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!eY4X!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png" width="1456" height="832" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:832,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1736206,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/216770255?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!eY4X!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 424w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 848w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 1272w, https://substackcdn.com/image/fetch/$s_!eY4X!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F602feada-9aa0-424c-975a-c9387c5be8d8_1659x948.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-GK1x3M" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-GK1x3M" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">The deadline begins while the evidence is still moving through the company.</figcaption></figure></div><p>Consider a Friday-afternoon report: a product-security analyst receives credible evidence that someone has exploited a vulnerability in a component shipped inside one of the company&#8217;s products.</p><p>The analyst doesn&#8217;t yet know which versions are affected. The incident team hasn&#8217;t established whether customer data was exposed. The product owner is in another time zone. Legal hasn&#8217;t seen the report. Nobody has confirmed which legal entity is the manufacturer responsible for the product in the European Union.</p><p>The reporting clock may already be running. But exploitation of that component elsewhere doesn&#8217;t, by itself, establish a reporting duty for this manufacturer.</p><p>Since September 11, 2026, manufacturers covered by the EU Cyber Resilience Act have had to report actively exploited vulnerabilities in their products and severe incidents affecting product security. The first warning is due without undue delay and no later than 24 hours after awareness. A fuller notification follows without undue delay and no later than 72 hours after that same awareness point. The final report comes later, after the manufacturer has had more time to investigate and respond.</p><p>That sequence matters. The law doesn&#8217;t wait for the incident team to finish the investigation before the company says something. It separates early warning from fuller assessment, so regulators receive notice while the facts are still developing.</p><p>For security leaders, the hard part isn&#8217;t writing the first report. It&#8217;s getting a fragmented organization to recognize that the clock has started.</p><div><hr></div><h2>TL;DR</h2><ul><li><p>The EU&#8217;s 24-hour warning requirement doesn&#8217;t wait for a completed investigation.</p></li><li><p>Compliance depends on a defined internal route from the person who first sees reliable evidence to the person authorized to report it.</p></li><li><p>Teams need a disciplined way to separate confirmed facts, current assessments and open questions.</p></li><li><p>The useful readiness test is whether the company can produce a defensible early warning on a Friday afternoon, not whether it has a finished policy.</p></li></ul><div><hr></div><h2>The first report is due while the facts are moving</h2><p>The Cyber Resilience Act covers hardware and software products made available on the EU market, including some components. Its main cybersecurity requirements don&#8217;t apply until December 11, 2027, but the reporting duties began on September 11, 2026.</p><p><a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX%3A32024R2847">The European Commission describes three stages.</a> An early warning is due within 24 hours after awareness. The fuller notification is due within 72 hours of that same awareness point. For an actively exploited vulnerability, a final report is due within 14 days after a corrective or mitigating measure becomes available. For a severe incident, the final report is due within one month after the manufacturer submits the 72-hour notification. In either case, a final report isn&#8217;t required if the relevant information has already been provided.</p><p>Each stage asks for a different level of detail. The early warning alerts regulators while the investigation is still developing. The 72-hour incident notification includes an initial assessment. The vulnerability notification asks for available details of the vulnerability and exploitation, including any corrective or mitigating measures. Final reports add the required impact and remediation details.</p><p>That&#8217;s a more honest model than pretending a company knows everything on day one. It&#8217;s also operationally unforgiving. An organization that treats reporting as the last step of an investigation will discover that the law treats it as one of the first.</p><p>The mistake to avoid is turning the 24-hour warning into a miniature final report. If every statement must survive the standard applied to a completed investigation, the team will spend the first day debating language while the deadline expires.</p><p>The opposite mistake is just as dangerous. Speed doesn&#8217;t excuse speculation. The early report needs a clean boundary between what the company knows, what it currently assesses and what it hasn&#8217;t established.</p><h2>A company learns through a person</h2><p>&#8220;The company became aware&#8221; sounds like a single event. It rarely is.</p><p>A customer tells support that a product behaved strangely. A researcher emails a security address. A threat-intelligence analyst connects an exploit to a library used in several products. A business unit receives the evidence, but the central incident team doesn&#8217;t see it until the next morning.</p><p>Which moment starts the clock?</p><p><a href="https://ec.europa.eu/newsroom/dae/redirection/document/131456">Commission guidance</a> says manufacturers should assess a suspicious event immediately. Awareness follows when that initial assessment gives them a reasonable degree of certainty that a vulnerability in their product is being actively exploited or that a severe incident has compromised the product&#8217;s security. The legal answer depends on the facts and should be worked through with counsel. The management problem is already visible: evidence can enter through many doors, while the duty belongs to the manufacturer as an organization.</p><p>Internal escalation has to begin before the reportability decision. Making that decision may require technical, product and legal coordination. That&#8217;s different from saying the statutory clock starts at the first signal.</p><p>The safer design is a low-friction escalation route for credible evidence of active exploitation or a potentially severe product-security incident. The first recipient doesn&#8217;t need authority to decide whether the CRA requires a report. That person needs to know where to send the evidence, what information to preserve and how to document the handoff.</p><p>Keep separate timestamps for receipt of evidence, the initial assessment and the point when the awareness threshold is met. Record who received what, when they received it and why it appeared credible enough to escalate. That doesn&#8217;t settle the legal question. It gives counsel and the response team a factual timeline instead of a reconstruction assembled two days later from chat messages and memory.</p><h2>The portal exposes the handoff</h2><p>The reporting mechanism itself reveals where organizations are likely to fail.</p><p>Manufacturers submit through ENISA&#8217;s Single Reporting Platform. They designate one primary Assigned Representative and can add up to 20 secondary representatives. Those people authenticate through EU Login and submit or update notifications for the organization. The primary representative needs a verified manufacturer association before inviting secondary representatives. Pending validation doesn&#8217;t itself prevent an initial notification.</p><p>That sounds administrative until the incident arrives. The platform launched without an API, so notifications must be submitted through its interface. Internal preparation can still be automated. Draft reports are stored in the individual representative&#8217;s account and aren&#8217;t visible to the organization&#8217;s other representatives. If the wrong national computer security incident response team is selected, the notification can be invalidated and must then be submitted again.</p><p>None of those details is the central policy question. Together, they show that the reporting path has human seams.</p><p>Someone must know which legal entity is the manufacturer. Someone must identify the correct national reporting destination. The default route is the CSIRT for the manufacturer&#8217;s main EU establishment; manufacturers without one follow fallback rules that can depend on an authorized representative, importer, distributor or user population. Someone must have a working account and multifactor authentication. Someone must collect the latest facts without overwriting the distinction between evidence and inference. Someone else must cover when the primary representative is unavailable.</p><p>This resembles the problem in <a href="https://firstorder.zenone.org/p/you-cant-patch-a-supply-chain-you">You Can&#8217;t Patch a Supply Chain You Can&#8217;t Inventory</a>. Product security depends on knowing which components sit inside which products. Regulatory reporting requires another map: which legal entity is the manufacturer, where product-security decisions are made and who can submit a report for that entity.</p><p>An escalation procedure that ends with &#8220;notify Legal&#8221; isn&#8217;t a reporting workflow. It&#8217;s a direction of travel.</p><h2>Separate facts from assessments and unknowns</h2><p>The first day&#8217;s work should produce a timely report that&#8217;s accurate about what&#8217;s known and explicit about what isn&#8217;t.</p><p>One simple discipline is to divide the working record into three parts:</p><p><strong>Confirmed facts.</strong> What evidence has the company received or reproduced? Which product, component or version is implicated? What time did the company receive the evidence?</p><p><strong>Current assessment.</strong> Why does the team believe the vulnerability is being actively exploited, or the incident may be severe? What scope and impact appear plausible based on the evidence available now?</p><p><strong>Open questions.</strong> Which versions, customers, data or functions remain under investigation? What will the team do next, and when will the assessment be updated?</p><p>This structure protects speed without pretending uncertainty has disappeared. It also makes later corrections easier to explain. A changed assessment isn&#8217;t necessarily a contradiction if the first report was explicit about what remained unknown.</p><p>External notification still carries consequences. As I argued in <a href="https://firstorder.zenone.org/p/one-way-doors">One-Way Doors</a>, disclosure can be an irreversible act made under pressure. Confidential regulatory notification isn&#8217;t the same as public disclosure. The response isn&#8217;t to wait for perfect knowledge. It&#8217;s to document how the decision was made and define who can authorize the submission.</p><h2>Run the Friday-afternoon test</h2><p>Policies written around calm working hours hide the weakness. Test the system when people and facts are scattered.</p><p>Choose one product sold in the EU. At 4:18 on a Friday afternoon, give a product-security analyst reliable evidence that attackers are exploiting a third-party component inside it. Don&#8217;t begin with the central incident commander. Begin where a real report might arrive.</p><p>Then watch the handoffs.</p><p>Can the first recipient identify the escalation route without searching a policy library? Can the company determine which products and versions contain the component? Can it identify the manufacturer and the correct national reporting destination? Can the assigned representative access the platform? Is there a backup? Can the team preserve the evidence and assessment timeline while counsel evaluates when the awareness threshold was met? Can it produce an early warning that distinguishes facts, assessments and unknowns?</p><p>Continue the exercise through the 72-hour notification. The question isn&#8217;t whether the company can solve the incident in three days. It&#8217;s whether the reporting record becomes more precise as the investigation advances.</p><p>The exercise should leave concrete repairs: a missing owner, an expired login, a product that isn&#8217;t mapped to a legal entity, an escalation address nobody monitors or a draft that only one person can see.</p><p>The reporting requirement is new, but the leadership problem isn&#8217;t. A deadline attached to organizational awareness forces the company to confront the delay between receiving evidence and authorizing action.</p><p>The 24-hour clock doesn&#8217;t require omniscience. It requires a company that can act before the investigation is complete.</p><h2>Sources</h2><ul><li><p><a href="https://digital-strategy.ec.europa.eu/en/policies/cra-reporting">European Commission: Cyber Resilience Act reporting obligations</a></p></li><li><p><a href="https://www.enisa.europa.eu/news/the-cra-single-reporting-platform-is-launched">ENISA: The CRA Single Reporting Platform is launched</a></p></li><li><p><a href="https://www.enisa.europa.eu/topics/product-security/vulnerability-services/eu-incident-response-and-cyber-crisis-management/single-reporting-platform-srp/frequently-asked-questions">ENISA: Single Reporting Platform frequently asked questions</a></p></li><li><p><a href="https://digital-strategy.ec.europa.eu/en/policies/cra-manufacturers">European Commission: CRA obligations for manufacturers</a></p></li></ul><p><em>Source note: The reporting stages, deadlines and portal mechanics in this article come from European Commission and ENISA guidance current as of September 21, 2026. Organizations should confirm scope and legal interpretation with counsel.</em></p><p><em>Analytical note: The argument that CRA readiness is an organizational handoff problem is my interpretation of those requirements and the platform&#8217;s operating model.</em></p>]]></content:encoded></item><item><title><![CDATA[Your Threat Model Still Assumes Human Attacker Labor]]></title><description><![CDATA[The attacks are familiar. AI is cutting the human work behind them, so some risks you've labeled 'low likelihood' need another look.]]></description><link>https://firstorder.zenone.org/p/your-threat-model-still-assumes-human</link><guid isPermaLink="false">https://firstorder.zenone.org/p/your-threat-model-still-assumes-human</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 15 Sep 2026 15:31:15 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!vHxV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Anthropic says the attacks in its September cases were familiar, but AI changed what those attacks cost to run by letting models handle reconnaissance, exploitation, tool development and data processing in parallel.</p></li><li><p>Google separately observed a mass credential-harvesting campaign planned, built and executed in under six hours.</p></li><li><p>Any accepted risk that depends on an attacker finding the work too manual, too specific to your environment or not valuable enough to pursue deserves another look.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!vHxV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!vHxV!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!vHxV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1955676,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/215601728?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!vHxV!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!vHxV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F831978f3-d723-4d24-83ba-de8a5d8137f3_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">Some risk decisions were priced around the hours an attacker would have to spend.</figcaption></figure></div><p>Some risks on your register are accepted because the control is good enough. Others are accepted because the attacker would have to work too hard. We rarely write that second sentence down.</p><p>Anthropic&#8217;s September 10 threat-intelligence report gives us a reason to write it down now. The attacks it investigated were mostly familiar: stolen credentials, unpatched internet-facing devices, exposed services, SQL injection and phishing. No new magic trick was required. What changed was the cost of running the campaign.</p><p>Anthropic reports breaches completed in two to three hours and dozens of victims handled in parallel by individual operators. It describes AI models taking over parts of reconnaissance, exploitation, tool development and stolen-data processing inside automated attack tooling. One actor used an AI-assisted workflow that automatically rebuilt and redeployed tooling after security products detected it.</p><p>Google reported a different case two days earlier. In one Q2 case, a financially motivated actor compromised a cloud resource, then used multiple AI agents working together to plan, build and execute a mass credential-harvesting campaign in under six hours. The system handled scanning, troubleshooting and changing network addresses with much less human intervention than a traditional attack would require.</p><div class="callout-block" data-callout="true"><p>The vulnerability didn't have to get worse. The labor got cheaper.</p></div><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading First Order! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><h2>The Attacks Didn&#8217;t Have to Get Exotic</h2><p>Security programs slow attackers down in more ways than we usually count. Recon takes time. Figuring out an unusual environment takes time. Chaining several ordinary weaknesses takes time. Rebuilding tooling after detection takes time. Sorting a pile of stolen data until someone knows what is worth selling or extorting takes time.</p><p>None of those delays is a formal security control by itself. They still affect which targets attackers choose and how long they&#8217;re willing to keep working on one.</p><p>Attackers have budgets too: people, access, compute, attention and hours in the day. A system can be technically exploitable and still be unattractive because the path is too manual, too weird or not valuable enough to justify specialist attention. That has always been part of the math, even when the risk register hid it inside a word like &#8220;unlikely.&#8221;</p><h2>Human Labor Was Hiding Inside &#8220;Low Likelihood&#8221;</h2><p>Look through enough risk-acceptance records and the notes explaining why the remaining risk is considered acceptable, and you&#8217;ll find versions of the same assumptions.</p><p>Exploitation requires substantial manual effort. The environment is unusual enough that broad targeting is unlikely. An attacker would have to chain too many weaknesses. Detection will force them to retool. The asset is not valuable enough to justify that much work.</p><p>Those can be rational reasons to rank one risk below another. They are also assumptions about attacker cost.</p><p>In May, I wrote about a narrower version of this in <a href="https://morphic.substack.com/p/the-day-the-clock-broke">The Day the Clock Broke</a>. That piece was about the economics of finding and weaponizing a logic flaw that requires understanding what the software is supposed to do. The September evidence is broader because discovery is not the only part getting cheaper.</p><p>Recon can run in parallel. Tooling can be adapted by a model. Errors can be retried without a human babysitting each step. Stolen data can be organized faster than one person could process it manually. The tooling can remember what it has already tried across sessions while other parts of the campaign keep running. The risk record may still say &#8220;requires a skilled attacker.&#8221; That may still be true, but the amount of skilled human time behind it may be much smaller.</p><h2>The Attacker&#8217;s Cost Changed</h2><p>Anthropic frames this in terms of attacker unit economics: how much it costs to run a campaign compared with what the attacker might gain. The potential payoff from stealing a customer database didn&#8217;t suddenly increase because AI entered the workflow. The cost of getting that payoff fell. A campaign can require less human labor, less specialist effort for some tasks and less elapsed time because more work can happen in parallel. That makes more targets worth the effort.</p><p>That doesn&#8217;t make every small company a priority target or every obscure system instantly exploitable. It means some targets that once required more effort than they were worth can become attractive without anything about the target itself changing.</p><p>The same logic applies to controls that slow an attacker down without stopping the attack outright. Detection still matters. Blocking still matters. Forcing an attacker to change tools still matters. But if part of a control&#8217;s value came from making a person spend hours diagnosing a failure, rewriting code and redeploying it, the control may consume less attacker time than it did when you designed it.</p><p>Anthropic saw exactly that loop in one case: security products detected the attacker&#8217;s tooling, then the AI-assisted workflow modified those tools and redeployed them. The detection still created work. The difference is that the operator no longer had to do all of it by hand.</p><p>Google&#8217;s under-six-hour credential campaign gives us a separate example. The attacker turned one compromised cloud resource into infrastructure for the campaign. The AI agents handled much of the scanning and troubleshooting, then the campaign collected credentials across a large number of targets.</p><h2>Don&#8217;t Mistake Case Studies for the Norm</h2><p>There is an important limit here. Anthropic says the cases in its report are notable and novel examples, not a picture of typical misuse across all cybercrime. Google says it hasn&#8217;t yet observed fully autonomous exploitation pipelines being used against real-world targets. Humans still make important decisions around target selection, monetization and review.</p><p>So I wouldn&#8217;t turn either report into &#8220;most attacks are agentic now.&#8221; The evidence doesn&#8217;t support that, and it doesn&#8217;t need to.</p><p>If you accepted a risk because exploitation looked too expensive to do repeatedly, you don&#8217;t need most attackers to adopt these workflows before the assumption deserves review. You need enough evidence that the time and skill required to run the attack have fallen. These two reports are enough to revisit a risk acceptance that depended on high attacker effort.</p><h2>Reprice the Risks Built on &#8220;Not Worth It&#8221;</h2><p>Instead of rebuilding every threat model around &#8220;AI attacks,&#8221; pull the accepted risks where the logic was basically &#8220;too tedious, too specific to our environment or not worth the attacker&#8217;s time.&#8221; Revisit those first.</p><p>Look for language like:</p><ul><li><p>exploitation requires substantial manual effort</p></li><li><p>the environment is too unusual to attract broad targeting</p></li><li><p>an attacker probably will not spend enough time chaining these weaknesses</p></li><li><p>repeated detection and retooling will make the campaign uneconomic</p></li><li><p>the asset is not valuable enough to justify specialist attention</p></li></ul><p>Examples: an old admin interface that&#8217;s reachable but unusual, several medium-severity weaknesses that only become dangerous when combined, or a detection control that forces an attacker to rebuild their tools every time.</p><p>Then ask what part of the likelihood rating depends on a person doing the work one target at a time.</p><p>Don&#8217;t automatically raise the impact score. The data, business process or system may be exactly as valuable as it was last quarter. Revisit the likelihood assumptions and any controls you counted on mainly because they slow the attacker down.</p><p>Start with the accepted risks that would hurt most if they happened. For each one, mark which attacker tasks you assumed a person would have to do, which controls mainly slow the attack rather than stop it and which systems you considered safer because they were obscure or unusual. You don&#8217;t need to reopen the entire risk register. Start with the risks you accepted partly because attacking them looked like too much work.</p><p>That&#8217;s also why I wouldn&#8217;t make this an &#8220;AI security&#8221; side exercise owned by the AI team. The affected risks may sit in identity, legacy infrastructure, SaaS, cloud, internet-facing systems or ordinary web applications. AI is changing the attacker&#8217;s cost, not the category of system being attacked.</p><p>Pick one accepted risk this week and read why you accepted it, not just its severity score. If the logic depends on the attacker deciding the work isn&#8217;t worth it, rerun the math.</p><p>The vulnerability may be exactly the same. The price is not.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/your-threat-model-still-assumes-human/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/your-threat-model-still-assumes-human/comments"><span>Leave a comment</span></a></p><div><hr></div><h3>Resources</h3><ul><li><p>Anthropic, &#8220;Detecting and countering misuse of AI: September 2026,&#8221; September 10, 2026: <a href="https://www.anthropic.com/threat-intelligence-report-september-2026">https://www.anthropic.com/threat-intelligence-report-september-2026</a></p></li><li><p>Google Threat Intelligence Group, &#8220;GTIG AI Threat Tracker: From Prompting to Autonomy - The Evolution of Adversarial AI,&#8221; September 8, 2026: <a href="https://cloud.google.com/blog/topics/threat-intelligence/from-prompting-to-autonomy-the-evolution-of-adversarial-ai">https://cloud.google.com/blog/topics/threat-intelligence/from-prompting-to-autonomy-the-evolution-of-adversarial-ai</a></p></li></ul><p><em>Source note: Anthropic explicitly says the cases in its report are notable and novel examples rather than typical misuse. Google says it has not yet observed fully autonomous exploitation pipelines operating against targets in the wild. This article treats the reports as evidence that attacker capability and operating cost are changing, not as a prevalence estimate for all cybercrime.</em></p><p><em>Analytical note: the argument that some accepted cyber risk embeds assumptions about attacker labor cost is my interpretation of the evidence above. Neither Anthropic nor Google audited enterprise risk registers or measured how often labor assumptions drive formal risk acceptance.</em></p>]]></content:encoded></item><item><title><![CDATA[Your AI Agent Can Commit the Company]]></title><description><![CDATA[Before security can enforce the limits, a business owner has to decide which discounts an agent can offer, payments it can send, exceptions it can approve and promises it can make.]]></description><link>https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company</link><guid isPermaLink="false">https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 08 Sep 2026 16:30:54 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!EPNv!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>In McKinsey&#8217;s August 25 survey, 40% of respondents at companies with more than $1 billion in annual revenue said their organizations were scaling AI agents, up from 27% last year. In Deloitte&#8217;s separate 2026 survey, only 21% of respondents said their organizations had mature governance for those agents.</p></li><li><p>Agent identity, action logs and policy checks while the agent is running matter. Those controls still can&#8217;t decide how much an agent may discount, refund or pay, or what it may promise.</p></li><li><p>The business owner sets the decision boundary. Security and engineering make it enforceable.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!EPNv!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!EPNv!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!EPNv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/cf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1835348,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/214233939?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!EPNv!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!EPNv!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fcf72d415-bf03-4357-8afd-a55e632ba4ba_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-GK1x3M" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-GK1x3M" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">An AI agent seeks authorization at a decision gate while a business owner sets the limit. The question isn&#8217;t what the agent can access. It&#8217;s what it can commit the company to.</figcaption></figure></div><p>Imagine a sales agent with legitimate access to the CRM, the company&#8217;s approved prices and discounts, and email.</p><p>A customer asks for a 12% discount. The agent sees that it may automatically approve discounts up to 15%, sends the revised quote and promises delivery by the end of the month. No security alert fires. Nobody stole a credential. The agent did exactly what the company&#8217;s rules allowed.</p><p>The harder question isn&#8217;t whether the agent had access to the pricing system.</p><p>Who decided the agent could approve a 15% discount without asking a person? Who decided it could promise a delivery date too?</p><p>I think a lot of companies are about to encounter that governance problem after their security controls work exactly as designed.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><p>McKinsey&#8217;s August 25 global survey found that 40% of respondents at companies with more than $1 billion in annual revenue said their organizations were scaling AI agents, up from 27% last year. In a separate Deloitte survey of 3,235 business and IT leaders across 24 countries, only 21% said their organizations had mature governance for those agents.</p><p>Those are different surveys with different samples, so I wouldn&#8217;t subtract one number from the other and call the difference a governance gap. But together they show the order in which many companies are moving: agents are entering real workflows before the companies deploying them have finished deciding what those agents should be allowed to do.</p><h2>The Business Decision Comes Before the Permission</h2><p>NIST is already working on the technical identity side. Its February concept paper separates four questions: Which agent is acting? What is it allowed to do? Can its actions be audited? Can the company later prove that the action came from that agent? OWASP&#8217;s Agent Control Standard, released September 1, adds checkpoints where agent platforms can inspect an action and enforce policy while the agent is running.</p><p>Those are useful controls. They don&#8217;t tell a sales leader whether 15% is the right discount ceiling.</p><p>They shouldn&#8217;t.</p><p>A policy engine can block a discount above 15%. It can&#8217;t decide whether 15% is commercially acceptable. The same problem applies to refund limits, payment thresholds, customer-policy exceptions and promises that consume real delivery capacity.</p><p>Security can ask whether the rule exists, whether the agent can bypass it and whether the action really stops when human approval is required. The business owner still has to set the number.</p><h2>Somebody Has to Fill In the Question Mark</h2><p>I wrote about the human version of this in <a href="https://firstorder.substack.com/p/the-delegation-illusion">The Delegation Illusion</a>. Real delegation gives another person more than a task. It defines which decisions they may make without asking, the limits on those decisions and what they should do when an unexpected case appears.</p><p>Agents need the same clarity in a form software can enforce.</p><p>With a person, missing criteria often surface as a question. &#8220;Can I go to 12%?&#8221; &#8220;Can I issue the credit?&#8221; &#8220;Do you want me to commit to Friday?&#8221; The conversation itself exposes the missing boundary.</p><p>An agent can keep moving.</p><p>Eventually somebody has to fill in a very boring line of configuration:</p><pre><code><code>max_discount_percent = ?</code></code></pre><p>The question mark isn&#8217;t an engineering problem.</p><p>Automated transactions aren&#8217;t entirely new legal territory. Massachusetts law, for example, says electronic agents can form a contract even when no person was aware of or reviewed their actions or the resulting terms. The statute also says the contract remains governed by the applicable substantive law, meaning the ordinary legal rules for that contract still apply.</p><p>That doesn&#8217;t mean every message from a modern AI agent creates an enforceable obligation for every company. These laws use &#8220;electronic agent&#8221; as an older term for automated software, and the legal effect of a particular action still depends on the facts and applicable law. But the rule removes one comforting assumption: a company can&#8217;t assume there was no commitment merely because no person reviewed the action.</p><p>Modern agents make that limit harder to define and enforce. They can interpret a broad goal, choose which tools to use and write the quote, email or contract language that communicates the decision outside the company.</p><h2>Payments Show How This Can Work</h2><p>The Agent Payments Protocol, or AP2, separates the agent making a purchase from the permission that authorizes the purchase.</p><p>In AP2&#8217;s human-not-present mode, the user approves an &#8220;open mandate&#8221; that records the limits for a future purchase. The user can then leave. When the agent selects a specific purchase, it creates and signs a &#8220;closed mandate&#8221; for that transaction. The receiving system can compare the purchase with the limits the user approved earlier.</p><p>The important part for this article isn&#8217;t the cryptography. It&#8217;s the sequence.</p><p>The person decides the boundary first. The agent acts inside it later. The receiving system has something concrete to verify.</p><p>Visa&#8217;s Trusted Agent Protocol approaches AI-assisted shopping from a different direction. It lets a merchant verify the agent&#8217;s identity, confirm that it is acting for a particular user and check that the user authorized the shopping task. The merchant can then limit the agent to that task.</p><p>Neither protocol decides how much the user should be willing to spend. They let software record the user&#8217;s limit and check a later transaction against it.</p><p>Companies need to follow the same order: the business owner sets the limit, security and engineering encode it, then the agent acts inside it.</p><h2>Security Shouldn&#8217;t Become the Business Owner</h2><p>There&#8217;s a predictable failure mode here for CISOs.</p><p>The company deploys an agent that can send quotes, move money, approve exceptions or make commitments to customers. Security is asked to put guardrails around it. During implementation, someone turns to the security team and asks what the threshold should be.</p><p>That&#8217;s the moment to give the question back.</p><p>A CISO should be able to answer five technical questions. Does the agent have its own identity? Do its credentials allow only the access it needs? Are important actions checked while it is running? Are its logs stored somewhere the agent can&#8217;t quietly rewrite them? Can the company revoke its access quickly? Security should challenge any design that has no approval threshold or no way to enforce one.</p><p>The CISO shouldn&#8217;t quietly become the person who decides how much margin sales may give away or how much credit support can issue. The same goes for which new vendors procurement may add or which service promise operations can make without checking capacity.</p><p>The owner will vary by company. Sales may decide which discounts an agent can offer. Finance may decide how much credit it can issue or money it can move. Procurement or legal may decide which vendors it can add and which contract terms it can accept. Operations may decide which delivery promises it can make. The org chart doesn&#8217;t have to look the same everywhere. But each limit has to come from the person already accountable for the result.</p><p>Then security and engineering can turn that decision into something the agent can&#8217;t casually route around.</p><h2>The Audit Trail Is Arriving First</h2><p>Congress is already focusing on verification and action logs.</p><p>On September 3, Representatives Josh Gottheimer and Mike Lawler announced the bipartisan Stop Rogue AI Act. It&#8217;s a proposal, not enacted law. The bill would direct NIST to develop standards and best practices for deploying AI agents securely. Those standards would cover continuous checks of agent actions, security and reliability testing, tamper-proof action logs and an inventory that other software can read.</p><p>All useful.</p><p>A perfect log can still record an undefined policy perfectly.</p><p>If an agent issued a $700 refund, the log can tell you which agent did it, which credential it used and when the call executed. It can&#8217;t tell you whether $700 should have been inside the agent&#8217;s authority unless the company made that decision before the action.</p><p>Companies often start with monitoring because monitoring tools are tangible. They can buy them, configure them and put the results on a dashboard. But those tools can&#8217;t decide where the agent&#8217;s autonomy should stop. A business owner has to set that limit, accept responsibility for it and revisit it when the business changes.</p><div class="pullquote"><p>The log is evidence that the boundary was followed. It&#8217;s not the boundary.</p></div><h2>Before the Next Agent Goes Live</h2><p>Pick one agent your company is deploying. Before reviewing its prompt, look at what the agent can actually do.</p><p>List every action the agent can take that commits the company to another party or materially affects a customer, company finances, operations or legal obligations. For each action, name the business owner who can set the limit. Define the exact point where the agent must stop and ask a person for approval. Then identify which system enforces that stop and who can change or remove it.</p><p>If you can&#8217;t name the business owner, don&#8217;t ask security to invent the limit.</p><p>If the threshold is &#8220;whatever the agent decides is reasonable,&#8221; you haven&#8217;t delegated a decision. You have delegated the creation of the decision rule too.</p><p>AI agents are getting capable enough that &#8220;what can it access?&#8221; is no longer a complete governance question. Ask what the agent can commit the company to, then ask who authorized that kind of action.</p><p>Only after that answer exists should somebody turn it into code.</p><div><hr></div><h3>Resources</h3><ul><li><p>McKinsey &amp; Company, &#8220;The state of AI in 2026: On the road to ROI,&#8221; August 25, 2026: <a href="https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai">https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai</a></p></li><li><p>Deloitte, &#8220;Business and IT leaders report AI agents are scaling faster than their guardrails,&#8221; April 24, 2026: <a href="https://www.deloitte.com/us/en/insights/topics/emerging-technologies/ai-agents-scaling-faster.html">https://www.deloitte.com/us/en/insights/topics/emerging-technologies/ai-agents-scaling-faster.html</a></p></li><li><p>NIST NCCoE, &#8220;Accelerating the Adoption of Software and Artificial Intelligence Agent Identity and Authorization,&#8221; February 5, 2026: <a href="https://csrc.nist.gov/pubs/other/2026/02/05/accelerating-the-adoption-of-software-and-ai-agent/ipd">https://csrc.nist.gov/pubs/other/2026/02/05/accelerating-the-adoption-of-software-and-ai-agent/ipd</a></p></li><li><p>OWASP GenAI Security Project, &#8220;Agent Control Standard,&#8221; September 1, 2026: <a href="https://genai.owasp.org/resource/agent-control-standard-acs/">https://genai.owasp.org/resource/agent-control-standard-acs/</a></p></li><li><p>Massachusetts General Laws, Chapter 110G, Section 14, &#8220;Automated transaction&#8221;: <a href="https://malegislature.gov/Laws/GeneralLaws/PartI/TitleXV/Chapter110G/Section14">https://malegislature.gov/Laws/GeneralLaws/PartI/TitleXV/Chapter110G/Section14</a></p></li><li><p>Agent Payments Protocol, AP2 specification and autonomous mandate flow: <a href="https://ap2-protocol.org/ap2/specification/">https://ap2-protocol.org/ap2/specification/</a></p></li><li><p>Visa, &#8220;Trusted Agent Protocol&#8221;: <a href="https://developer.visa.com/use-cases/trusted-agent-protocol">https://developer.visa.com/use-cases/trusted-agent-protocol</a></p></li><li><p>Office of Congressman Mike Lawler, &#8220;Exclusive: New bill cracks down on AI agents after Hugging Face breach,&#8221; September 3, 2026: <a href="https://lawler.house.gov/news/documentsingle.aspx?DocumentID=6424">https://lawler.house.gov/news/documentsingle.aspx?DocumentID=6424</a></p></li></ul><p><em>Legal note: the automated-transaction example above is illustrative, not legal advice. Whether a specific agent action creates an enforceable obligation depends on the applicable law, the facts and the authority surrounding that transaction.</em></p><p><em>Analytical note: the distinction between a business owner&#8217;s decision and the technical permission given to an agent is my interpretation of the sources above. NIST, OWASP, AP2 and Visa describe ways to identify agents, authorize actions, enforce rules while an agent is running and verify a user&#8217;s intent. They don&#8217;t decide who inside a company has authority to set the business limits.</em></p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/your-ai-agent-can-commit-the-company/comments"><span>Leave a comment</span></a></p>]]></content:encoded></item><item><title><![CDATA[You Can't Patch a Supply Chain You Can't Inventory]]></title><description><![CDATA[The White House's new grid-security order is about foreign hardware. The harder lesson is what happens when security risk gets installed years before anyone can replace it.]]></description><link>https://firstorder.zenone.org/p/you-cant-patch-a-supply-chain-you</link><guid isPermaLink="false">https://firstorder.zenone.org/p/you-cant-patch-a-supply-chain-you</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 01 Sep 2026 16:30:35 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!wFgo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>The White House declared a national emergency over certain foreign-produced equipment in the U.S. bulk-power system, including risks from remote access, software, firmware and maintenance dependencies.</p></li><li><p>The order doesn&#8217;t prove hidden backdoors exist throughout the grid. It does show how hard remediation becomes after critical hardware is installed.</p></li><li><p>Supply-chain security starts with knowing what you own, how it communicates and what replacing it would actually take.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!wFgo!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!wFgo!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!wFgo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/fd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:2318081,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/213583784?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!wFgo!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!wFgo!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ffd4a1ce5-f17c-451f-ba8c-4e57d17e1535_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">The grid&#8217;s attack surface was procured years ago.</figcaption></figure></div><p>On August 26, the White House declared a national emergency over the foreign supply of equipment used in the U.S. bulk-power system.</p><p>The phrase that got the attention was &#8220;digital backdoors.&#8221; The order says some foreign-produced equipment might contain them and could allow remote foreign access. It doesn&#8217;t say investigators proved that hidden backdoors exist throughout the U.S. grid.</p><p>That distinction matters because a credible risk isn&#8217;t the same thing as a confirmed compromise.</p><p>The more useful word appears later in the order: <strong>inventory</strong>.</p><p>The Department of Energy is directed to identify risky equipment and recommend ways to identify, inventory, isolate, monitor or replace it. For equipment that is already installed, the Secretary can impose conditions that include securing, disconnecting, replacing or removing it.</p><p>That sequence should look familiar to every security leader. Before you can contain something, you have to know it&#8217;s there.</p><h2>The Hardware Is Already There</h2><p>The new order covers much more than transformers. It reaches grid-connected inverters, battery energy-storage systems, generators, protective relays, industrial control systems, programmable logic controllers, intelligent electronic devices and associated software, firmware, remote-access capabilities and lifecycle maintenance mechanisms.</p><p>Some of that equipment will be installed after the order. Some of it is already operating.</p><p>The latter category is harder.</p><p>Before the government directs isolation, disconnection, replacement or removal of existing equipment, the order requires Energy to consider reliability, safety, the availability of secure replacements and continuity of essential service. That&#8217;s not ceremonial caution. It&#8217;s the operating constraint.</p><div class="pullquote"><p>You can&#8217;t treat a power transformer like a compromised laptop.</p></div><p>DOE&#8217;s 2024 report to Congress on large power transformers describes them as expensive, highly customized and difficult to transport. It says current manufacturing lead times can reach and exceed 36 months. DOE said again this month that critical grid equipment can carry lead times of two years or more.</p><p>So imagine the security answer is &#8220;replace it.&#8221; Fine. With what? How quickly? What does the grid do while you wait? Can the replacement be transported to the site? Is there another supplier that can meet the electrical and physical requirements?</p><p>The vulnerability may be technical. The remediation is constrained by physics, manufacturing capacity and procurement decisions made years earlier.</p><h2>The Firewall Assumed the Hardware Was Telling the Truth</h2><p>There&#8217;s another reason this story deserves more than a procurement-policy reading.</p><p>In May 2025, Reuters reported that U.S. experts had found undocumented communications equipment inside some Chinese-made solar inverters. One of its two anonymous sources also said undocumented communications devices, including cellular radios, had been found in some batteries from multiple Chinese suppliers.</p><p>The caveats are important. Reuters couldn&#8217;t determine how many inverters or batteries had been examined. The sources weren&#8217;t named. The report didn&#8217;t establish that every Chinese inverter contained undocumented hardware, that the devices had been used maliciously or that the U.S. grid had been broadly compromised.</p><p>Still, the mechanism matters.</p><p>Utilities can put firewalls around the communications paths they know about. An undocumented radio creates a path the network design may not know exists. The firewall didn&#8217;t suddenly fail. The inventory was incomplete.</p><p>That&#8217;s a different class of security problem. The documented interface may not be the complete interface.</p><p>A software vulnerability often gives you something concrete to hunt for: a version, a hash, a package, a CVE. Hardware and firmware supply-chain questions can be uglier. What is actually inside the device? Which components came from which suppliers? Which maintenance channel can reach it? What behavior is undocumented because nobody knew to ask?</p><p>The Reuters report isn&#8217;t proof of a hidden kill switch. It&#8217;s evidence that provenance and interface discovery deserve a place in the threat model.</p><h2>Procurement Is Architecture</h2><p>Security teams often inherit supply-chain risk after the consequential decisions have already been made.</p><p>The vendor was selected. The contract was signed. The equipment was ordered. The remote-support model was accepted. The update path was designed. The dependency on the manufacturer became part of operations.</p><p>Then security gets asked whether the system is safe.</p><p>NIST&#8217;s new July 2026 due-diligence guide for cybersecurity supply-chain risk management is useful here because it refuses to reduce the problem to a vendor questionnaire. Its assessment components include foreign ownership, control or influence, provenance, resilience, foundational cyber practices and supply-chain tiers.</p><p>FERC has been moving in the same direction for the bulk-power system. In September 2025 it approved new action on supply-chain risk-management reliability standards, building on requirements for entities to address risks in industrial-control-system hardware, software and services.</p><p>The common thread is visibility before dependence.</p><p>Procurement determines more of your eventual security architecture than most organizations admit. It can determine who can maintain the equipment remotely, how updates arrive, which supplier dependencies sit underneath the product and whether you have a realistic replacement path when the risk changes.</p><p>By the time an incident responder sees an alert, those choices may already be sunk costs.</p><h2>The Flag on the Box Is Not a Control</h2><p>The executive order explicitly pushes federal procurement toward U.S.-manufactured energy infrastructure. That can reduce some geopolitical and supply-disruption risks. It doesn&#8217;t turn country of origin into a complete security model.</p><p>NIST&#8217;s July guidance makes that clear without making a political argument. Foreign ownership, control or influence is one component of due diligence. So are provenance, resilience, basic cyber practices and the tiers behind the supplier you are directly buying from.</p><p>Domestic equipment can still have vulnerable firmware, weak remote access or fragile supplier dependencies. Foreign equipment isn&#8217;t one homogeneous risk category either.</p><p>So I&#8217;d resist the easy translation of this order into &#8220;buy American and the security problem goes away.&#8221; The government itself is building a more granular model. The order allows the Secretary of Energy to establish criteria for pre-qualified equipment and vendors, while also evaluating particular transactions and existing equipment based on risk.</p><p>The useful question isn&#8217;t which flag is printed on the shipping crate.</p><p>It&#8217;s whether you understand the equipment&#8217;s provenance, communications paths, update and maintenance mechanisms, supplier dependencies and replacement options well enough to make a decision when the threat model changes.</p><p>The grid makes this problem dramatic because the equipment is enormous, the service is essential and some replacements take years. The leadership lesson travels well beyond utilities.</p><p>Take the technology your organization can&#8217;t operate without and look at three things: remote-access or maintenance channels you don&#8217;t directly control, supplier or component tiers you can&#8217;t currently trace and replacements that would take months or years instead of days.</p><p>If the answer is &#8220;we don&#8217;t know,&#8221; that doesn&#8217;t mean the equipment is compromised.</p><p>It means you don&#8217;t know which options you have if a vulnerability, sanctions decision, supplier failure or incident suddenly changes the acceptable risk.</p><p>That&#8217;s why inventory belongs at the beginning of this story, not after the headline about backdoors.</p><p>In 2020, DOE was already restricting certain Chinese-supplied bulk-power equipment serving critical defense facilities. In 2025, FERC tightened the supply-chain conversation again. Now the federal government has declared a national emergency and is explicitly considering how to identify, inventory, isolate, monitor and replace risky equipment already in the field.</p><p>Two facts can sit next to each other without much editorial help: the risk has been on the government&#8217;s radar for years, and the installed base still matters enough to require a new emergency order.</p><p>Security debt isn&#8217;t always old code.</p><p>Sometimes it arrived on a pallet, passed procurement and became part of the infrastructure before anyone asked how hard it would be to remove.</p><p>If a critical system can&#8217;t be inventoried, isolated, monitored or replaced without threatening the business it supports, procurement already made part of your incident-response plan.</p><p>Security should know what it bought before the incident decides for you.</p><div><hr></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><h3>Resources</h3><ul><li><p>Executive Order 14421, <em>Declaring a National Emergency to Secure the United States Bulk-Power System</em>, August 26, 2026, published in the Federal Register August 31, 2026: <a href="https://www.govinfo.gov/content/pkg/FR-2026-08-31/pdf/2026-17843.pdf">https://www.govinfo.gov/content/pkg/FR-2026-08-31/pdf/2026-17843.pdf</a></p></li><li><p>The White House, <em>Fact Sheet: President Donald J. Trump Declares a National Emergency to Secure America&#8217;s Bulk-Power System</em>, August 26, 2026: <a href="https://www.whitehouse.gov/fact-sheets/2026/08/fact-sheet-president-donald-j-trump-declares-a-national-emergency-to-secure-americas-bulk-power-system/">https://www.whitehouse.gov/fact-sheets/2026/08/fact-sheet-president-donald-j-trump-declares-a-national-emergency-to-secure-americas-bulk-power-system/</a></p></li><li><p>U.S. Department of Energy, <em>Large Power Transformer Resilience Report to Congress</em>, July 2024: <a href="https://www.energy.gov/sites/default/files/2024-10/EXEC-2022-001242%20-%20Large%20Power%20Transformer%20Resilience%20Report%20signed%20by%20Secretary%20Granholm%20on%207-10-24.pdf">https://www.energy.gov/sites/default/files/2024-10/EXEC-2022-001242</a> </p></li><li><p>U.S. Department of Energy, <em>Strengthening America&#8217;s Grid Supply Chain</em>, August 2026: <a href="https://www.energy.gov/oe/articles/strengthening-americas-grid-supply-chain">https://www.energy.gov/oe/articles/strengthening-americas-grid-supply-chain</a></p></li><li><p>Reuters, <em>Rogue communication devices found in Chinese solar power inverters</em>, May 14, 2025, republished by <a href="http://investing.com/">Investing.com</a>: <a href="https://www.investing.com/news/stock-market-news/ghost-in-the-machine-rogue-communication-devices-found-in-chinese-inverters-4043741">https://www.investing.com/news/stock-market-news/ghost-in-the-machine-rogue-communication-devices-found-in-chinese-inverters-4043741</a></p></li><li><p>Federal Energy Regulatory Commission, <em>FERC Takes Action to Enhance Reliability of the U.S. Electric Grid</em>, September 18, 2025: <a href="https://www.ferc.gov/news-events/news/ferc-takes-action-enhance-reliability-us-electric-grid">https://www.ferc.gov/news-events/news/ferc-takes-action-enhance-reliability-us-electric-grid</a></p></li><li><p>National Institute of Standards and Technology, SP 1326, <em>Cybersecurity Supply Chain Risk Management: Due Diligence Assessment Quick-Start Guide</em>, July 8, 2026: <a href="https://csrc.nist.gov/pubs/sp/1326/final">https://csrc.nist.gov/pubs/sp/1326/final</a></p></li><li><p>U.S. Department of Energy, <em>Secretary of Energy Signs Order to Mitigate Security Risks to the Nation&#8217;s Electric Grid</em>, December 17, 2020: <a href="https://www.energy.gov/articles/secretary-energy-signs-order-mitigate-security-risks-nations-electric-grid">https://www.energy.gov/articles/secretary-energy-signs-order-mitigate-security-risks-nations-electric-grid</a></p></li></ul><p><em>Source note: the Federal Register identifies the August 26 order as Executive Order 14421. The White House order page currently labels it 14420, which duplicates the number shown for an August 10 order on the White House executive-order index. I use the Federal Register number here and avoid relying on the White House page for the identifier.</em></p><p><em>Analytical note: the argument that procurement decisions become part of incident-response capability is my interpretation of the order, DOE replacement constraints, NIST due-diligence guidance and FERC supply-chain requirements. The Reuters findings are reported findings based on anonymous sources and an undisclosed sample size, not proof of broad grid compromise.</em></p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading First Order! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div>]]></content:encoded></item><item><title><![CDATA[The Board Can't Read a Risk Appetite You Never Set]]></title><description><![CDATA[Many boards run cyber risk reporting without ever defining a cyber risk appetite. NACD guidance and a 2026 CISO survey show why the missing decision matters.]]></description><link>https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 25 Aug 2026 16:31:16 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!AEDH!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Pulse Security AI&#8217;s 42-person survey found 71% spent 10+ hours preparing each board cycle while 55% said their board had no formally defined cyber risk appetite.</p></li><li><p>NACD&#8217;s 2026 guidance says the board and management should define that appetite, then report current exposure against it.</p></li><li><p>Better slides can&#8217;t replace the missing decision about how much risk the company is willing to carry.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!AEDH!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!AEDH!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 424w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 848w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 1272w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!AEDH!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png" width="1456" height="971" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/f417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:971,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1935817,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/212628798?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!AEDH!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 424w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 848w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 1272w, https://substackcdn.com/image/fetch/$s_!AEDH!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff417e7e9-9e63-4dbf-ac00-7bfa429b18a6_1536x1024.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">The reading can be precise and still tell you nothing about what the company is willing to accept.</figcaption></figure></div><p>Pulse Security AI surveyed 42 security leaders and corporate directors about board reporting.</p><p>Seventy-one percent said they spend at least ten hours preparing for each board cycle. Fifty-five percent said the board has never formally defined the company&#8217;s cyber risk appetite.</p><p>Those two numbers don&#8217;t belong in separate conversations.</p><p>The first says security teams spend a lot of time building the report. The second says many of those reports have no agreed threshold to report against.</p><p>The usual advice is better storytelling. Shorten the deck. Replace CVSS scores with business scenarios. Add financial exposure. Put the architecture diagram in the appendix.</p><p>And that&#8217;s all useful.</p><p>But eventually somebody has to answer a different question: <strong>How much of this risk are we willing to carry?</strong></p><p>If nobody has answered it, the CISO can spend two days polishing the presentation and still walk into the room without the one thing that makes the colors mean anything.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><h2>Ten Hours Against No Baseline</h2><p>The Pulse numbers need the usual vendor-survey caveat.</p><p>Pulse is a security-program-management company. Its survey had 42 respondents, 70% of them CISOs or heads of security, plus more than 20 in-depth interviews and two workshops involving roughly 22 CISOs. Pulse says directly that the statistics aren&#8217;t nationally representative.</p><p>So I wouldn&#8217;t turn 55% into an estimate of every corporate board in America.</p><p>I&#8216;d treat it as a field signal.</p><p>Especially because the 2026 NACD and Internet Security Alliance <em>Director&#8217;s Handbook on Cyber-Risk Oversight</em> describes exactly the governance step the survey says is often missing.</p><p>NACD&#8217;s Principle Four says the board should work with management to define the organization&#8217;s cyber-risk appetite: the amount and types of risk it is willing to accept while pursuing its objectives. NACD says that appetite should be explicit, quantitative where possible and expressed in business or financial terms. Management then operates the program inside that direction.</p><p>Principle Five completes the loop. NACD says board reporting should show current exposure against the board-approved risk appetite.</p><div class="pullquote"><p>First define the boundary. Then report where you are relative to it.</p></div><p>That&#8217;s a different job from making a technical dashboard easier to understand.</p><h2>Red and Green Need a Reference Point</h2><p>A critical vulnerability can be technically severe whether your board likes it or not. Severity and business acceptance are different questions.</p><p>The board isn&#8217;t there to decide whether CVSS 9.8 is a large number. It needs enough information to understand what the scenario can do to the business, how exposed the company is and whether management is operating inside the risk boundary the company agreed to.</p><p>That last part disappears when the appetite was never defined.</p><p>Imagine two identical red boxes on two board dashboards.</p><p>At Company A, red means estimated exposure exceeds a board-approved threshold and management needs a decision on treatment.</p><p>At Company B, red means the security team thinks the situation is bad.</p><p>Same color. Different governance.</p><p>Company A encoded a prior decision into the report. Company B encoded an opinion.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><p>This is where I think we sometimes give CISOs the wrong communication advice. We tell them to stop speaking security and start speaking business. Fine. But business language doesn&#8217;t create a business decision nobody made.</p><p>Pulse found another version of this in its small sample. Forty-nine percent primarily described cyber-risk severity with qualitative categories. Only 3% primarily used quantified dollar amounts.</p><p>I wouldn&#8217;t turn that into &#8220;everything needs a dollar sign.&#8221; Fake precision is not maturity. NACD itself says quantitative <strong>where possible</strong>. A qualitative boundary can still be a boundary.</p><p>The failure is having none.</p><h2>&#8220;Speak Business&#8221; Is Downstream Advice</h2><p>Take ransomware against a critical operation.</p><p>A CISO can explain the likely interruption, which services are exposed, what recovery looks like, which controls are missing and the plausible financial consequences.</p><p>The board may understand every word.</p><p>There&#8216;s still another decision: is that exposure acceptable?</p><p>Maybe the company is willing to carry more risk because reducing it would delay a major launch. Maybe the potential loss sits outside anything the company wants to tolerate. Maybe management should transfer part of it through insurance, spend to reduce it or explicitly accept it for six months.</p><p>Security can model those choices and recommend one. Security shouldn&#8217;t quietly make the enterprise risk decision by deciding whether a box is red.</p><p>NIST IR 8286A Rev. 1 comes at the same problem from the enterprise-risk side. The December 2025 revision describes cybersecurity risk analysis in the context of risk appetite and risk tolerance, then uses risk registers to help decision-makers prioritize risk response and monitoring.</p><p>So the threshold is not decoration around the risk analysis. It It&#8217;s an input to it.</p><p>If the company never established that input, the CISO ends up reverse-engineering risk appetite from whatever gets approved, delayed or funded.</p><p>You find out what the company was supposedly willing to risk only after somebody reacts to the recommendation.</p><p>That&#8217;s backwards.</p><h2>The Board Meeting Should Produce a Decision</h2><p>Pulse reported one number I think deserves more attention than it got: half of respondents said their boards had made no explicit decision to accept, mitigate or transfer cyber risk during the prior year.</p><p>Again, small sample.</p><p>Still worth asking about your own board.</p><p>If a quarterly cyber meeting produces questions, discussion and another request for slides but never an explicit risk decision, what exactly is the meeting governing?</p><p>My read is that a lot of organizations have turned the cyber board deck into a briefing instrument. The CISO explains the environment. Directors ask questions. Everybody leaves better informed.</p><p>NACD&#8217;s model asks more of the meeting.</p><p>The report should show where current exposure sits relative to a boundary the board already approved. When exposure crosses that boundary, management reduces it, transfers it, avoids it or comes back with a reason the company should accept more.</p><p>That changes the package.</p><p>For each material scenario, I&#8217;d want five things: the scenario, current exposure, the agreed appetite or tolerance, management&#8217;s recommended response and the decision required from the board.</p><p>Examples: ransomware disrupting a core operation, compromise of sensitive customer data, failure of a critical third party.</p><p>Then stop.</p><p>The vulnerability inventory can live somewhere else.</p><p>And if the company has never defined the appetite, don&#8217;t hide that absence behind a heat map. Put it on the agenda.</p><h2>The Missing Decision</h2><p>Only 12.5% of the security leaders in Pulse&#8217;s survey said they were very confident that the board walked away from a presentation understanding the true state of the security program.</p><p>The obvious response is another round of communication training.</p><p>Maybe some CISOs need it.</p><p>But communication isn&#8217;t always the first failure.</p><p>A board report can explain the security posture clearly and still leave directors unable to judge whether the posture is acceptable. That happens when reporting gets built before risk appetite does.</p><p>So before spending another ten hours on the deck, look at the red, yellow and green boxes and ask what each one is measured against.</p><p>If the answer is a board-approved risk boundary, good. Improve the report.</p><p>If the answer is &#8220;we haven&#8217;t actually set one,&#8221; stop polishing slide four.</p><p>That&#8217;s the first item on the agenda.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-board-cant-read-a-risk-appetite/comments"><span>Leave a comment</span></a></p><div><hr></div><h3>Resources</h3><ul><li><p>Pulse Security AI, <em><a href="https://pulsesecurity.ai/newsroom/ciso-board-communication-gap/">The CISO-Board Communication Gap</a></em>, August 2026. Survey of 42 security leaders and corporate directors, 20+ interviews and two moderated workshops with roughly 22 CISOs. Pulse explicitly states that the statistics are not nationally representative.</p></li><li><p>National Association of Corporate Directors and Internet Security Alliance, <em><a href="https://www.nacdonline.org/all-governance/governance-resources/governance-research/director-handbooks/2026-cyber-risk-oversight/cyber-risk-handbook-principles-2026/principle-4-adopt-an-enterprise-framework-for-managing-cyber-risk/">2026 Director&#8217;s Handbook on Cyber-Risk Oversight, Principle Four: Adopt an Enterprise Framework for Managing Cyber Risk</a></em>, April 16, 2026.</p></li><li><p>National Association of Corporate Directors and Internet Security Alliance, <em><a href="https://www.nacdonline.org/all-governance/governance-resources/governance-research/director-handbooks/2026-cyber-risk-oversight/cyber-risk-handbook-principles-2026/principle-5-guide-cybersecurity-risk-measurement-reporting/">Principle Five: Guide Cybersecurity Risk Measurement and Reporting</a></em>, April 16, 2026.</p></li><li><p>NIST IR 8286A Rev. 1, <em><a href="https://csrc.nist.gov/pubs/ir/8286/a/r1/final">Identifying and Estimating Cybersecurity Risk for Enterprise Risk Management</a></em>, December 2025.</p></li></ul><p><em>Analytical note: the argument that an undefined risk appetite can make otherwise accurate board reporting less decision-useful is my interpretation of the governance guidance and survey findings above. Pulse&#8217;s sample is small and non-representative.</em></p>]]></content:encoded></item><item><title><![CDATA[The 51-Point Gap: Your People Got Faster. The Work Didn't.]]></title><description><![CDATA[Workers report strong personal productivity gains from AI while organizational work changes much less. In security, the bottleneck is moving into judgment.]]></description><link>https://firstorder.zenone.org/p/the-51-point-gap-your-people-got</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-51-point-gap-your-people-got</guid><pubDate>Tue, 18 Aug 2026 16:32:17 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!JEdd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Gallup&#8217;s May 2026 data says 65% of employees at organizations using AI report better personal productivity. Only 14% strongly agree AI has changed how work gets done. The 51 points in the title come from subtracting the second number from the first.</p></li><li><p>ISC2 found 65% of cybersecurity professionals surveyed are spending more time deciding whether to trust or act on AI recommendations, while 63% are spending more time checking the output.</p></li><li><p>AI can shorten the first part of the job without shortening the whole job. Review, ownership and human judgment can become the new delay.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!JEdd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!JEdd!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!JEdd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/bcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:1973084,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.zenone.org/i/211584736?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!JEdd!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 424w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 848w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 1272w, https://substackcdn.com/image/fetch/$s_!JEdd!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbcf60f5e-5f87-498f-9ae0-859d273dcc8e_1672x941.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">The answer got there fast. The decision still has to make the trip.</figcaption></figure></div><p>People are reporting that AI makes them faster. The organizations around them seem to be changing much more slowly.</p><p>Gallup&#8217;s current workplace AI indicator puts numbers on that difference. Among employees at U.S. organizations that have implemented AI, 65% say it has had a positive effect on their productivity and efficiency. By comparison, 14% strongly agree AI has transformed how work gets done in their organization. That&#8217;s the 51-point spread I&#8217;m writing about: 65% minus 14%.</p><p>I&#8217;m doing that subtraction, not Gallup. The 51 points aren&#8217;t a formal Gallup metric, and the two questions use different response scales. So I wouldn&#8217;t treat 51 as some new productivity index. It&#8217;s a useful signal that the individual experience and the organizational experience may be moving at very different speeds.</p><p>Gallup&#8217;s Q2 numbers make that worth watching. Forty-seven percent of U.S. employees say their organization has integrated AI tools, up from 41% the previous quarter. Fifty-two percent say they use AI in their role at least occasionally, and 30% use it a few times a week or more.</p><p>People are using the tools. Now follow what happens to their work after the tool finishes.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><h2>Faster Worker, Same System</h2><p>Take a security analyst using AI to cut a first-pass investigation from twenty minutes to five. Great. Then the recommendation waits for senior review. Maybe an application owner has to weigh in. Maybe somebody needs to approve a change or verify the evidence before anyone acts.</p><div class="pullquote"><p>The analyst saved fifteen minutes. The incident may not have.</p></div><p>Security already knows this pattern. Add a lot more finding capacity without adding triage capacity and you don&#8217;t magically get fewer vulnerabilities. You can get a bigger queue. AI can create the same problem in ordinary workflows: more useful output arriving faster at the same people who were already responsible for checking it.</p><p>Gallup doesn&#8217;t measure that downstream effect. I&#8217;m making the connection. But Microsoft&#8217;s 2026 Work Trend Index gives us another reason to look there. Its global survey covered 20,000 AI users across 10 markets. Only 26% said leadership was clearly and consistently aligned on AI. Microsoft also found that organizational factors such as culture, manager support and talent practices accounted for 67% of the statistical association with reported AI impact, compared with 32% for individual factors such as mindset and behavior.</p><p>Microsoft is careful here, and we should be too. That&#8217;s an association, not proof of cause. But it points at something leaders can inspect inside their own companies. Someone can get very good at using AI while their work still passes through the same approvals, handoffs and decision rules it did two years ago.</p><p>The tool changed. The route didn&#8217;t.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-51-point-gap-your-people-got?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-51-point-gap-your-people-got?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><h2>Security Is Already Feeling It</h2><p>ISC2 surveyed 856 cybersecurity professionals who use AI in their roles. Sixty-five percent said they were spending more time deciding when to trust or act on AI-generated recommendations. Sixty-three percent were spending more time reviewing or validating AI output.</p><p>That&#8217;s the part I care about.</p><p>AI can remove effort from one part of the job and put some of it back somewhere else. The work didn&#8217;t disappear. Some of it moved into deciding whether the answer is good enough to act on.</p><p>ISC2 also found that 48% were spending less time on tasks that don&#8217;t involve AI assistance. At the same time, half of respondents said human decision-makers bear final accountability when an AI-recommended action turns out to be wrong. Nearly a quarter said they&#8217;re often or very often expected to act on AI-generated security output without fully understanding how it was produced.</p><p>So the machine can produce the recommendation faster, while the person sitting behind it still owns the mistake.</p><p>That changes the job.</p><p>SANS saw something related in its July AI survey. AI use in cybersecurity rose from 50% to 78% in a year, yet only 27% of practitioners described their deployments as mature production. Seventy-six percent said security now has a governance role for enterprise AI. Sixty-three percent reported significant shortcomings in AI-assisted threat detection and response.</p><p>The tools arrived quickly. The rules for trusting them, checking them and deciding who owns the result are taking longer.</p><h2>The Bottleneck Moved</h2><p>Security teams have spent years automating the slow parts of the job. AI speeds up another chunk of it. But if the output still sits waiting for review, approval or someone willing to own the decision, you haven&#8217;t removed the delay. You&#8217;ve moved it.</p><p>An alert might be summarized in seconds and then wait an hour because nobody trusts the summary enough to close it. Or AI might rank a vulnerability instantly, only for the ticket to sit because the application owner isn&#8217;t clear. The machine has already finished. The work hasn&#8217;t.</p><p>That distinction matters because a lot of AI programs are still being measured at the tool. Seats. Usage. Prompts. Agents deployed.</p><p>Those numbers tell you whether people are using AI. They don&#8217;t tell you whether anything finishes sooner, let alone with accuracy.</p><p>If you want to know that, measure what happens after the output appears. How long until somebody checks it? How often does it get sent back? How often does a human reverse the recommendation? Most important, when does someone finally act?</p><p>If the first step went from twenty minutes to five and the final action still takes three hours, AI probably did what you asked it to do.</p><p>You found the next constraint.</p><h2>Run Two Clocks</h2><p>Pick one security workflow and run two clocks.</p><p>Stop the first when AI produces something a competent person considers useful. Stop the second when the organization actually does something with it.</p><p>The distance between those clocks is where I&#8217;d look next. Maybe the work is sitting with a reviewer. Maybe nobody knows who can approve it. Maybe the evidence isn&#8217;t strong enough yet. Whatever the reason, you can finally see where the time went.</p><p>Alert triage is an obvious place to try this because the start and finish are relatively easy to see. But the test works anywhere AI touches a workflow. Don&#8217;t ask only whether the analyst got faster. Ask whether the decision happened sooner and whether the result was still good.</p><p>I wouldn&#8217;t turn this into a staffing argument yet. Some tasks will genuinely require fewer human hours. Others may create more review because AI lets people produce plausible-looking work much faster than before. You have to measure the whole path before you know which one you have.</p><p>Gallup&#8217;s 51-point spread doesn&#8217;t prove organizations have failed to change. It tells us employees report a much bigger change in their own productivity than they report in how work gets done around them.</p><p>That difference is worth looking more deeply at.</p><p>If AI gives your analyst an answer in thirty seconds and your organization still needs three days to act on it, stop measuring the thirty seconds.</p><p>Follow the work downstream.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-51-point-gap-your-people-got/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-51-point-gap-your-people-got/comments"><span>Leave a comment</span></a></p><div><hr></div><h3>Resources</h3><ul><li><p>Gallup, <a href="https://www.gallup.com/699797/indicator-artificial-intelligence.aspx">Artificial Intelligence Indicator</a>. May 2026 data: 65% positive self-reported productivity among employees at organizations that implemented AI, 14% strongly agree AI transformed how work gets done, plus survey methods and current adoption measures.</p></li><li><p>Gallup, <a href="https://www.gallup.com/workplace/712736/organizational-adoption-jumps-six-points.aspx">Organizational AI Adoption Jumps Six Points</a>, July 20, 2026. Reports organizational integration rising from 41% to 47% in Q2, 52% total workplace AI use and 30% frequent use.</p></li><li><p>Microsoft, <a href="https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization">2026 Work Trend Index: Agents, Human Agency, and the Opportunity for Every Organization</a>, May 5, 2026. Global survey of 20,000 AI users across 10 markets, including leadership alignment and the 67% vs. 32% organizational/individual association. Microsoft states the finding is associative, not causal.</p></li><li><p>ISC2, <a href="https://www.isc2.org/insights/2026/07/rethinking-ai-impact-on-cybersecurity-roles">Rethinking AI&#8217;s Impact on Cybersecurity Roles</a>, July 14, 2026. Survey of 856 cybersecurity professionals using AI, including changes in validation time, decision pressure and accountability.</p></li><li><p>SANS Institute, <a href="https://www.sans.org/press/announcements/ai-use-cybersecurity-jumped-from-50-to-78-year-ai-related-failures-rose-sharply-too-new-sans-institute-survey-reveals-governance-gap">AI Use in Cybersecurity Jumped From 50% to 78% in a Year</a>, July 13, 2026. Survey of 536 cybersecurity and IT practitioners plus a module of 57 senior security leaders, including maturity, governance and detection/response findings.</p></li></ul><p><em>The queue and two-clock examples are my interpretation of the survey results. Gallup, Microsoft, ISC2 and SANS didn&#8217;t measure those exact workflow effects.</em></p>]]></content:encoded></item><item><title><![CDATA[The Security Team You Cut Doesn't Come Back on Incident Day]]></title><description><![CDATA[Cyber capacity looks expensive in the quiet months. Then the pager goes off and the spreadsheet discovers what it removed.]]></description><link>https://firstorder.zenone.org/p/the-security-team-you-cut-doesnt</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-security-team-you-cut-doesnt</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Tue, 11 Aug 2026 16:31:38 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!tosg!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong>TL;DR</strong></p><ul><li><p>Security capacity is easiest to cut when nothing is visibly on fire, which can make preparedness look like unused labor.</p></li><li><p>Recent CISA workforce cuts and a later pledge to restore 600 positions make the problem unusually visible: cyber capability doesn&#8217;t rebuild at incident speed.</p></li><li><p>The better leadership test is not utilization. It&#8217;s whether the organization can absorb a serious incident without discovering that one exhausted person was the control.</p></li></ul><div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!tosg!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!tosg!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 424w, https://substackcdn.com/image/fetch/$s_!tosg!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 848w, https://substackcdn.com/image/fetch/$s_!tosg!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 1272w, https://substackcdn.com/image/fetch/$s_!tosg!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!tosg!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:909016,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/210671534?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!tosg!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 424w, https://substackcdn.com/image/fetch/$s_!tosg!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 848w, https://substackcdn.com/image/fetch/$s_!tosg!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 1272w, https://substackcdn.com/image/fetch/$s_!tosg!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9788780c-b765-4c45-9a1c-908b5ba05ed0_1600x900.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A cybersecurity responder sits at multiple monitoring screens in a dark operations room during an incident.</figcaption></figure></div><p>Security capacity looks wasteful when nothing is burning. The incident channel is quiet, the queue is manageable and a senior responder who isn&#8217;t actively responding to anything can start to look like expensive empty space on a staffing sheet. That&#8217;s usually when somebody gets clever with the math.</p><p>The role is consolidated. The second person who knows the ugly part of the identity stack is declared redundant. Regional coverage gets thinner. Training gets pushed because production work is measurable and preparedness mostly isn&#8217;t. On a normal Tuesday, the spreadsheet looks cleaner. Then the pager goes off, and the spreadsheet discovers what it removed.</p><h2>The Spreadsheet Sees Idle</h2><p>Cybersecurity has a measurement problem that often gets disguised as a staffing problem. We are very good at counting visible work: alerts closed, vulnerabilities patched, tickets handled and assessments completed. We are much worse at pricing the ability to handle the thing that hasn&#8217;t happened yet.</p><p>That ability can look suspiciously like slack. A senior responder who has enough room to think. Two people who can each perform a critical function even though only one is needed during normal operations. An engineer who remembers why the ugly exception exists because they were there when it was made. A regional advisor whose value sits partly inside relationships thatdon&#8217;t not generate tickets.</p><p>Put those people into a utilization model and some will look expensive. Put them into a real incident and the arithmetic changes quickly. NIST&#8217;s current incident-response guidance treats incident response as part of cybersecurity risk management across the organization, not as a task that begins after detection. If response capability has to exist before the event, some of what leadership is buying will look quiet before it looks valuable.</p><p>The hard part is that quiet capacity and useless capacity can look almost identical from far enough away. That&#8217;s exactly why a spreadsheet built around ordinary-week utilization is such a weak instrument for deciding what the organization can safely lose.</p><h2>CISA Is Running the Experiment in Public</h2><p>The current fight over the Cybersecurity and Infrastructure Security Agency makes the problem unusually visible. In June, Sen. Mark Warner said CISA had lost nearly one-third of its workforce since January 2025, primarily senior career officials. In May, Sen. Maggie Hassan separately wrote that CISA had lost more than a third of its workforce in 2025, including almost all of its senior leaders.</p><p>Those are statements from elected officials, not an independent census of the agency, so I wouldn&#8217;t treat either percentage as a sacred measurement. The direction is still difficult to miss. CISA lost substantial experienced capacity, and on June 28 Warner publicly welcomed what he described as CISA&#8217;s plan to rehire 600 federal employees.</p><p>In her May letter to CISA&#8217;s acting director, Hassan put it plainly: </p><blockquote><p>In 2025, for example, CISA lost more than a third of its workforce, including almost all its senior leaders, raising questions in the private sector and Congress about the direction of the agency.</p></blockquote><p>There&#8217;s a useful detail on the other side of the aisle too. In January, the Republican-led House Homeland Security Committee said its FY2026 DHS appropriations package included $20 million to strengthen the CISA workforce and directed the agency to maintain sufficient staffing, preserve regional field offices and ensure every state and territory had dedicated cybersecurity-advisor support.</p><p>The politics are different, but the operational fact underneath them is not: cyber capability has to be standing there before the bad day arrives. That doesn&#8217;t mean every cut is reckless or every security role is sacred. It means leadership should understand what a role is buying before deciding that today&#8217;s unused capacity is tomorrow&#8217;s safe savings.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-security-team-you-cut-doesnt?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-security-team-you-cut-doesnt?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><h2>Capacity Is More Than Headcount</h2><p>A lazy version of this argument would be &#8220;never cut security people.&#8221; I don&#8217;t buy that. Security organizations can get bloated, teams can keep doing work long after the reason disappeared and a large headcount can still produce a weak program.</p><p>Headcount is not capability, but capability is not headcount-free either. When an experienced person leaves, the organization can lose more than one line on an org chart. It may lose the memory of why a control was built a certain way, the person who knows which vendor actually answers after midnight, the responder who can separate familiar noise from the first five minutes of something ugly and the second pair of hands that lets the first responder sleep before hour twenty-six.</p><p>Some of that can be documented. Good teams document aggressively. Some of it lives in what people have seen, who they know and what they notice before they can fully explain why it feels wrong.</p><p>GAO&#8217;s July review of the Federal Rotational Cyber Workforce Program is a useful reminder that building this kind of depth is not automatic. GAO reported that 634 employees applied to the program over its life, while only eight were approved to serve rotations. It also found the program had effectively been halted after low participation and shifting priorities.</p><p>The point is not that rotations solve the workforce problem. The point is that judgment, cross-agency familiarity and operational depth take deliberate repetition to build. Once that depth leaves, a requisition is not the same thing as replacement.</p><h2>The Utilization Trap</h2><p>A security function run at permanent full utilization is a system with no shock absorber. It can look wonderfully efficient right up until something abnormal happens, then every person pulled into response creates another hole behind them. Vulnerability work pauses. Access reviews slip. Engineering projects stall. Someone who already owns two critical systems inherits a third because the one person who knew it best is now living in the incident bridge.</p><p>This is where utilization and readiness stop being the same thing. The corporate instinct is to ask whether people are busy enough. Security leadership has to ask something more useful: busy enough for which day?</p><p>A normal week and a serious incident are different operating modes. Designing the team entirely around the first and expecting it to absorb the second is like sizing emergency power around average grid demand. It works beautifully until the reason you bought emergency power finally shows up.</p><p>That is why I am suspicious of the phrase &#8220;extra capacity.&#8221; Extra compared with what: the average ticket queue, last quarter&#8217;s incident count or the worst event the company has already survived? A spare tire is underutilized right up until the interesting part of the trip.</p><h2>Run the Incident-Day Test</h2><p>Before cutting a security team, I would want a different sheet. Not names and salaries first. Capabilities. Take the credible incident you least want next quarter and run the staffing model against it.</p><p>Who is on the bridge in the first hour, and who can replace them six hours later? Which critical functions have one person who actually understands the machinery? What normal security work stops while response is underway? Which partner or vendor relationships disappear when one employee leaves?</p><p>Then look at the second day, because that is where the fantasy of unlimited human capacity usually gets expensive. Who slept? Who is making consequential decisions after sixteen hours awake? Who can communicate with executives while somebody else stays close enough to the technical detail to know whether the story is drifting away from reality?</p><p>There is a difference between redundancy and resilience. Redundancy is two people who can do the same job. Resilience is having enough depth, memory and relationships that the system can take a hit without turning one exhausted person into the architecture.</p><div class="pullquote"><p>Resilience is having enough depth, memory and relationships that the system can take a hit without turning one exhausted person into the architecture.</p></div><p>If the cuts still make sense after that test, make them. Security should not be exempt from hard choices merely because the work is scary. But the choice should be explicit: this is the capability we are reducing, this is the failure mode we are accepting and this is what we will not be able to recreate quickly if we guessed wrong.</p><p>That is a much cleaner conversation than arguing about benchmark headcount or pretending there is a universal ratio of security people to employees. A board does not need to know whether the team looks lean. It needs to know what the organization can still do when the incident consumes the people normally doing everything else.</p><p>That is the capacity number. If the answer depends on calling back the people you cut six months earlier, you did not save capacity. You spent it.</p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading First Order! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><div><hr></div><h3>Resources</h3><ul><li><p>U.S. Sen. Mark Warner, &#8220;Warner Raises Alarm on CISA Workforce and Budget Cuts That Are Leaving Our Country Vulnerable to Threats,&#8221; June 16, 2026. <a href="https://www.warner.senate.gov/newsroom/press-releases/warner-raises-alarm-on-cisa-workforce-and-budget-cuts-that-are-leaving-our-country-vulnerable-to-threats/">https://www.warner.senate.gov/newsroom/press-releases/warner-raises-alarm-on-cisa-workforce-and-budget-cuts-that-are-leaving-our-country-vulnerable-to-threats/</a></p></li><li><p>U.S. Sen. Mark Warner, &#8220;Senate Intel Vice Chair Warner Statement on Secretary Mullin&#8217;s Pledge to Rehire 600 Staffers,&#8221; June 28, 2026. <a href="https://www.warner.senate.gov/newsroom/press-releases/senate-intel-vice-chair-warner-statement-on-secretary-mullins-pledge-to-rehire-600-staffers/">https://www.warner.senate.gov/newsroom/press-releases/senate-intel-vice-chair-warner-statement-on-secretary-mullins-pledge-to-rehire-600-staffers/</a></p></li><li><p>U.S. Sen. Maggie Hassan, &#8220;Senator Hassan Presses for Answers on Major Reported Data Leak at Leading Cybersecurity Agency,&#8221; May 19, 2026. <a href="https://www.hassan.senate.gov/news/press-releases/senator-hassan-presses-for-answers-on-major-reported-data-leak-at-leading-cybersecurity-agency">https://www.hassan.senate.gov/news/press-releases/senator-hassan-presses-for-answers-on-major-reported-data-leak-at-leading-cybersecurity-agency</a></p></li><li><p>U.S. House Committee on Homeland Security, &#8220;Chairman Garbarino Celebrates House Passage of the FY2026 Department of Homeland Security Appropriations Bill,&#8221; January 22, 2026. <a href="https://homeland.house.gov/2026/01/22/chairman-garbarino-celebrates-house-passage-of-the-fy2026-department-of-homeland-security-appropriations-bill/">https://homeland.house.gov/2026/01/22/chairman-garbarino-celebrates-house-passage-of-the-fy2026-department-of-homeland-security-appropriations-bill/</a></p></li><li><p>U.S. GAO, &#8220;Cyber Workforce: Agencies Did Not Widely Adopt Rotational Program,&#8221; July 16, 2026. <a href="https://www.gao.gov/products/gao-26-108736">https://www.gao.gov/products/gao-26-108736</a></p></li><li><p>NIST, &#8220;Incident Response Recommendations and Considerations for Cybersecurity Risk Management: A CSF 2.0 Community Profile,&#8221; SP 800-61 Rev. 3, April 2025. <a href="https://csrc.nist.gov/pubs/sp/800/61/r3/final">https://csrc.nist.gov/pubs/sp/800/61/r3/final</a></p></li></ul>]]></content:encoded></item><item><title><![CDATA[Nobody Rents Floor Space to Retention]]></title><description><![CDATA[Only 34% of cybersecurity pros plan to stay, and a 54-point satisfaction gap shows why. What backing security teams reveals about retention and CISO burnout.]]></description><link>https://firstorder.zenone.org/p/nobody-rents-floor-space-to-retention</link><guid isPermaLink="false">https://firstorder.zenone.org/p/nobody-rents-floor-space-to-retention</guid><pubDate>Fri, 07 Aug 2026 17:15:44 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!-4YV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!-4YV!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!-4YV!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!-4YV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png" width="1456" height="813" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:813,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:8209213,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/210176488?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!-4YV!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!-4YV!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3edec906-f67f-4cb6-abc3-bb9108fd0574_2752x1536.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">An empty conference hall after the sessions are over. Rows of places built for gathering, with one chair already out of line.</figcaption></figure></div><p><em>Only 34% of security professionals surveyed plan to stay with their current employer. The sharper signal is a 54-point career-satisfaction gap tied to whether security feels like a core organizational priority.</em></p><p>I&#8217;m still in Las Vegas. Black Hat has wrapped and DEF CON is underway, which puts an unusual amount of the security profession within a few miles of each other for a week.</p><p>It&#8217;s a strange place to be thinking about retention.</p><p>Or maybe exactly the right one</p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading First Order! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><p>IANS Research and Artico Search surveyed more than 500 security professionals for their 2026 Cybersecurity Talent Report, published in April. Only <strong>34% said they plan to stay with their current employer</strong>.</p><p>That does not mean the other 66% have resignation letters drafted. Some are considering a move. Some are probably undecided. But only about a third are willing to say, yes, I plan to remain here.</p><p>For a field that spends so much time talking about talent shortages, that&#8217;s a pretty uncomfortable number.</p><h2>The 54-point number</h2><p>The same report contains a finding I think is more useful than the headline.</p><p>Among security staff who see security as a core organizational priority, <strong>73% report being satisfied with their careers</strong>. Among those who perceive little or no organizational backing for security, satisfaction falls to <strong>19%</strong>.</p><p>That&#8217;s a 54-point gap.</p><p>Not between the highest-paid people and the lowest-paid. Not between executives and analysts. The split in this particular finding is how respondents perceive their organization&#8217;s commitment to the work.</p><p>That&#8217;s correlation, and it&#8217;s worth saying so clearly. The survey does not prove that organizational backing causes career satisfaction. People who are happier may also judge their organizations more favorably, and there are probably other variables moving underneath both numbers.</p><p>Still, 73 to 19 is hard to wave away.</p><p>It suggests that what people believe about the organization&#8217;s commitment to security is connected to how they feel about the career they&#8217;re building inside it.</p><h2>The pay answer, and where it runs out</h2><p>Pay people more.</p><p>Fair answer.</p><p>Money matters. I&#8217;ve never met an engineer who works day after day for gratitude, and anyone trying to explain retention while pretending compensation is irrelevant has probably skipped an important part of the problem.</p><p>But the IANS and Artico data adds an interesting wrinkle: <strong>wage growth matters more for retention than absolute pay level</strong>. Employees receiving even modest increases reported higher satisfaction and stronger intent to stay than employees whose compensation remained flat, regardless of where the absolute number started.</p><p>That makes intuitive sense to me. The number matters, but movement carries information too.</p><p>A raise says something changed.</p><p>Steve Martano, an IANS faculty member and partner at Artico Search, framed the larger issue around rising expectations and constrained resources. His point was that top talent is looking for more than compensation alone. Visibility, career growth and support from security leadership matter too.</p><p>So I don&#8217;t think the data says pay doesn&#8217;t matter.</p><p>It says pay isn&#8217;t the whole message.</p><h2>What people are working inside</h2><p>The pressure underneath these retention numbers isn&#8217;t especially mysterious.</p><p>ISC2&#8217;s 2025 Cybersecurity Workforce Study surveyed 16,029 people responsible for cybersecurity around the world. <strong>59% reported critical or significant skills needs on their teams</strong>, up from 44% the year before.</p><p>More interesting to me is what happened downstream.</p><p>ISC2 reports that <strong>88% of respondents experienced at least one significant cybersecurity consequence because of a skills deficiency within their team or wider organization</strong>. Sixty-nine percent reported more than one.</p><p>That isn&#8217;t a prediction about what a shortage might eventually do.</p><p>Respondents were describing consequences they said had already happened.</p><p>You can pay somebody well and still place them inside a system where the responsibility keeps expanding faster than the resources available to meet it. The salary changes. The queue doesn&#8217;t. The person is still accountable for an outcome they increasingly suspect the organization has not equipped them to produce.</p><p>Eventually that becomes information.</p><p>Not just about the job.</p><p>About the place.</p><h2>The person expected to fix it may be looking too</h2><p>Leadership isn&#8217;t standing outside this problem.</p><p>Proofpoint&#8217;s 2025 Voice of the CISO surveyed 1,600 CISOs across 16 countries. <strong>Sixty-three percent said they had experienced or witnessed burnout during the previous year.</strong> The same report found 66% facing what respondents described as excessive expectations.</p><p>Then IANS and Artico surveyed 662 CISOs for their 2026 State of the CISO benchmark. <strong>Nearly seven in ten said they were open to making a career move within the next year.</strong></p><p>That doesn&#8217;t mean seven in ten are quitting either.</p><p>It does mean the people we&#8217;re asking to stabilize security teams are themselves unusually mobile.</p><p>I think that matters when we reduce retention to something a manager is supposed to fix with better one-on-ones.</p><p>The manager may be looking at the same door.</p><h2>What Vegas builds really well</h2><p>This week makes the contrast difficult to miss.</p><p>Black Hat&#8217;s Business Hall had more than 400 vendors and startups this year. Its own description talks about networking, making connections and discovering career opportunities alongside all the products, demos and meetings.</p><p>Then there&#8217;s BSides Las Vegas.</p><p>Its <strong>Hire Ground</strong> program has existed since at least 2016. The archived description talks openly about interview preparation, recruiters, resume help, social profiles and career mapping. Hire Ground is still on the BSidesLV schedule in 2026, with sessions this week on high-performing security teams, getting hired, surviving the cyber job market and standing out in it.</p><p>That&#8217;s useful infrastructure. I&#8217;m not criticizing it.</p><p>People need ways into better jobs. People need ways out of bad ones.</p><p>But we&#8217;ve gotten remarkably good at building machinery around movement.</p><p>Recruiting systems. Career tracks. Interview coaching. Resume reviews. Job boards. Search firms. Conference programming.</p><p>There&#8217;s a whole apparatus for helping somebody become somebody else&#8217;s employee.</p><p>I can&#8217;t find an equivalent apparatus for making the current place worth staying.</p><h2>Nobody rents floor space to retention</h2><p>That&#8217;s the line I keep coming back to.</p><p>Nobody rents floor space to keeping people.</p><p>There&#8217;s no retention booth because retention isn&#8217;t really a product. You can&#8217;t buy it in a hall at Mandalay Bay, integrate it next quarter and put a green check beside it.</p><p>It&#8217;s mostly ordinary decisions.</p><p>Whether someone gets the headcount they were told was coming. Whether the project that matters gets funded. Whether a strong engineer can see an actual next step instead of another year of &#8220;we&#8217;re working on the leveling framework.&#8221; Whether the security leader is allowed to make decisions or simply absorbs accountability from people who can.</p><p>None of that demos especially well.</p><p>It still shows up in the numbers.</p><h2>The mentorship answer</h2><p>Nick Kakolowski, IANS&#8217;s Senior Research Director, argues that leaders should invest harder in mentorship, coaching and career development. The idea is straightforward: give people a sense of progression and purpose before burnout gets to make the decision for them.</p><p>I agree with that.</p><p>I also think mentorship has a failure mode we don&#8217;t talk about enough.</p><div class="pullquote"><p>A mentor who can&#8217;t move can&#8217;t model movement.</p></div><p>If the person coaching you has no budget, no headcount, no authority over progression and no realistic path of their own, eventually the coaching becomes a description of a room neither person can change.</p><p>Good conversation. Same room.</p><p>Mentorship works when it&#8217;s attached to something real.</p><p>A decision someone can make. A training budget that actually exists. Scope that can expand. A title that can change. Compensation that moves. Access to work that develops the person instead of simply consuming them.</p><p>Otherwise we&#8217;ve handed somebody a sympathetic listener and called it a career path.</p><h2>What &#8220;backing security&#8221; actually looks like</h2><p>This is where I think the 73% versus 19% number gets interesting.</p><p>Saying security is important is free.</p><p>Every company can do that.</p><p>The team learns what you mean by watching everything that costs something.</p><p>Budgets. Hiring approvals. Deferred projects. Incident staffing. Whether security gets involved before or after a decision. Whether the CISO has access to the people making the tradeoffs. Whether the same three people keep absorbing the work created by every new priority.</p><p>Those are signals.</p><p>People are good at reading them.</p><p>Which may be why &#8220;organizational backing&#8221; shows up so loudly in the satisfaction data. Whatever the direction of causation, employees are not judging the company&#8217;s commitment from the values page.</p><p>They&#8217;re living inside the allocation decisions.</p><h2>The test I&#8217;d rather use</h2><p>I&#8217;ve never loved tenure by itself as a measure of a healthy team.</p><p>Long tenure can mean someone is thriving. It can also mean they haven&#8217;t looked recently, the market isn&#8217;t cooperating or leaving would cost too much personally.</p><p>I&#8217;d rather ask something harder:</p><div class="pullquote"><p><strong>Could this person leave without guilt, and would I help them do it well?</strong></p></div><p>A team where the answer is yes has a chance of making staying an actual choice.</p><p>That&#8217;s different from retention through friction.</p><p>Different from loyalty through dependency.</p><p>Different from quietly building a room with a difficult door.</p><p>And I think people can feel the difference.</p><h2>The hall for the other half</h2><p>Black Hat is over now. DEF CON is going.</p><p>Vegas is still full of movement: people meeting each other, exchanging ideas, discovering companies, finding opportunities and figuring out what they might do next.</p><p>That machinery works.</p><p>I&#8217;m glad it exists.</p><p>But after looking at the retention numbers, I&#8217;m more interested in the machinery that doesn&#8217;t have a hall.</p><p>The part where someone goes back to work after all of this and decides, again, that the place they&#8217;re already in is still worth choosing.</p><p>Nobody gets a booth for that.</p><p>Somebody still has to build it.</p><div class="captioned-button-wrap" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/nobody-rents-floor-space-to-retention?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;}" data-component-name="CaptionedButtonToDOM"><div class="preamble"><p class="cta-caption">Thanks for reading First Order! This post is public so feel free to share it.</p></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/nobody-rents-floor-space-to-retention?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/nobody-rents-floor-space-to-retention?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p></div><div><hr></div><h3>Resources</h3><ul><li><p><a href="https://www.iansresearch.com/resources/press-releases/detail/new-ians-and-artico-search-report-finds-only-34--of-cybersecurity-professionals-plan-to-stay-in-their-current-roles">2026 Cybersecurity Talent Report announcement - IANS Research and Artico Search</a></p><p>Primary source for the survey of 500+ security professionals, the 34% stay-intent figure, 73%/19% career-satisfaction split, wage-growth finding and comments from Nick Kakolowski and Steve Martano.</p></li><li><p><a href="https://www.isc2.org/insights/2025/12/2025-ISC2-Cybersecurity-Workforce-Study">2025 ISC2 Cybersecurity Workforce Study</a></p><p>Primary study based on 16,029 respondents. Source for the 59% critical/significant skills-needs figure and the 88%/69% reported security-consequence figures.</p></li><li><p><a href="https://www.proofpoint.com/us/newsroom/press-releases/proofpoint-2025-voice-ciso-report">Proofpoint 2025 Voice of the CISO</a></p><p>Primary source for the survey of 1,600 CISOs across 16 countries and the 63% experienced-or-witnessed-burnout figure.</p></li><li><p><a href="https://www.iansresearch.com/resources/press-releases/detail/2026-report-finds-executive-level-ciso-titles-more-prevalent-than-ever">2026 State of the CISO Benchmark - IANS Research and Artico Search</a></p><p>Primary source for the 662-CISO sample and the finding that nearly seven in ten CISOs are open to a career move within the next year.</p></li><li><p><a href="https://www.blackhat.com/us-26/business-hall.html">Black Hat USA 2026 Business Hall</a></p><p>Black Hat&#8217;s official description of the 2026 Business Hall, including its 400+ vendors and startups, networking focus and career-opportunity language.</p></li><li><p><a href="https://bsideslv.org/schedule">BSides Las Vegas 2026 Schedule</a></p><p>Official 2026 schedule showing Hire Ground and its current career-focused programming.</p></li><li><p><a href="https://archive.bsideslv.org/2016/hire-ground/">BSidesLV Hire Ground archive, 2016</a></p><p>Archived description of Hire Ground&#8217;s interview preparation, recruiter networking, resume support, social-profile help and career mapping.</p></li><li><p><a href="https://info.defcon.org/defcon34/menu/">DEF CON 34 Conference Information</a></p><p>Official DEF CON 34 conference information for the August 6-9, 2026 event in Las Vegas.</p></li></ul>]]></content:encoded></item><item><title><![CDATA[Elevation or Exposure]]></title><description><![CDATA[They say security leaders finally arrived: the strategic seat, AI governance, a CFO-sized job. Against the liability numbers, a lot of that arrival is a bigger blame surface.]]></description><link>https://firstorder.zenone.org/p/elevation-or-exposure</link><guid isPermaLink="false">https://firstorder.zenone.org/p/elevation-or-exposure</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Thu, 30 Jul 2026 17:32:57 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!2T3S!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!2T3S!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!2T3S!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 424w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 848w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 1272w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!2T3S!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp" width="1456" height="819" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:819,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:89498,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/webp&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/209008839?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!2T3S!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 424w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 848w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 1272w, https://substackcdn.com/image/fetch/$s_!2T3S!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1b26dad0-1e83-41af-9c0e-74d89189a43d_1456x819.webp 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A single terminal screen glows in a dim room. The cursor blinks on a dashboard showing zero alerts. The real activity happens elsewhere - in pipelines no one can trace. That&#8217;s the space where the money moves and the watchman sees nothing.</figcaption></figure></div><p>Someone offered me a bigger title once, and the thing I remember isn&#8217;t the title. It&#8217;s the list that came with it. Everything I&#8217;d now be responsible for, printed out, longer than the list before it. I kept waiting for the second page, the one that says what I&#8217;d be allowed to decide, and it wasn&#8217;t in the folder. Nobody hands you that page. You&#8217;re supposed to be too flattered to ask for it.</p><p>I&#8217;ve been thinking about that folder a lot lately, because the numbers say the entire security profession is being handed the same one. This April, Gravitee put a survey in front of 750 senior technology leaders across the UK and US. 54% reported an experienced or suspected AI agent security or data-privacy incident in the prior twelve months. Over a third confirmed one outright.</p><p>Now set the growth curve next to that number: the enterprise agent footprint roughly doubled between December 2025 and April 2026, and nearly 38% of organizations already run more than 100 agents in production. The incidents aren&#8217;t trailing adoption. They&#8217;re pacing it.</p><p>When one surfaces, the board often asks this question: why didn&#8217;t <em>you</em> see it?</p><p>Notice the pronoun. Not why didn&#8217;t <em>we</em> see it. <em>You</em>. Accountability had already landed on one chair, and nobody paused to ask whether that chair came with the access to answer.</p><p>AI governance has now been added to the job description, usually as one line: ensure AI safety. Splunk&#8217;s 2026 CISO Report asked 650 CISOs across nine countries, and 96% said they oversee AI governance and risk across the enterprise.</p><p>Fine. Look at what the mandate actually covers, though. Procurement runs agents that negotiate vendor terms and HR has them screening candidates and drafting offer letters. These systems don&#8217;t retrain themselves overnight, but they don&#8217;t hold still either. Configurations drift. Prompts get tuned. A vendor pushes an update and tool access quietly widens. The agent your team reviewed on Monday isn&#8217;t quite the agent making decisions on Friday and no change ticket marks the difference. Many were bought and wired in by teams the security org has never audited.</p><p>I call them shadow agents. Some were never approved. Others were approved once and changed underneath the approval. They&#8217;re shadows not because someone deliberately hid them, but because no one built a durable line of sight from deployment to consequence.</p><p>Forrester thinks they&#8217;re security&#8217;s problem now: its 2026 threat list ranks AI agents second, personal agents slipping in through browser hooks and inbox access, acting at machine speed outside governance, with CISOs accountable for exposure they have limited control over. And they don&#8217;t enter at the bottom of the org chart. TrustedTech Team surveyed 2,001 UK and US employees this spring: 65% of senior decision-makers use AI tools their employer hasn&#8217;t approved. The employees under them? 31%.</p><p>So here&#8217;s the ledger the board never reads. The mandate: clear. The tooling: partial and fragmented. The access: restricted, since the systems live under budget lines security doesn&#8217;t own. The authority to pull one out of production? Often zero.</p><p>Accountability moved into the seat. It doesn&#8217;t appear that anything else came with it.</p><p>The generous read says I&#8217;m describing a promotion, not a trap.</p><p>And it&#8217;s worth taking seriously, because smart people make it. Deloitte&#8217;s 2026 Global Technology Leadership Study surveyed more than 660 senior technology executives and put the shift in one line: &#8220;the era of the operational technologist is over.&#8221; The argument runs: security spent twenty years asking for a seat at the strategy table and AI governance is finally that seat. The CISO stops being the firewall keeper and becomes steward of the organization&#8217;s digital conscience. Not compliance anymore: judgment.</p><p>There&#8217;s real logic here. Somebody has to lead the security and assurance side of AI governance, and few executives in the building have spent a career studying how systems fail when a motivated person wants them to fail. Legal thinks in liability. Engineering thinks in uptime. Security already lives in adversaries and blast radius, which is exactly the muscle this moment demands.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><p>If the CISO can&#8217;t lead this, who can?</p><p>It&#8217;s a fair question.</p><p>So I ran the test it implies.</p><p>Pick one AI agent already in production. Something touching real money: loan underwriting, maybe payment routing. Then ask the executive who&#8217;s accountable for it to map that single agent&#8217;s lifecycle end to end: where the data comes from, what the agent can touch and who owns the decision when the agent gets it wrong. And somewhere in that map, make them show you the gate between recommendation and action. For consequential decisions, the model proposes; something more deterministic - a policy engine, a validation check or, where the stakes require it, a human - disposes.</p><p>My bet on where it breaks: the surface answers arrive fast. Vendor name, model family, the liability clause in the contract. Then you ask what the agent actually saw before last Tuesday&#8217;s decision, what context and instructions shaped its recommendation, and what authorized the system to act on it. The trail dies at the first integration boundary. Not because anyone&#8217;s hiding it... because nobody built the layer that would show it. Opaque to the buyer and often to the builder too.</p><p>If the ownership and evidence actually exist, a credible first-pass map shouldn&#8217;t require an excavation. I doubt most organizations could produce that map today. Gravitee&#8217;s numbers suggest that isn&#8217;t just my experience: mean monitoring coverage sits near 52%, leaving roughly half of production agents outside security&#8217;s line of sight.</p><p>Stewardship of a system you can&#8217;t inspect isn&#8217;t stewardship. It&#8217;s faith.</p><p>The trap is the assumption underneath the title: visibility tells you what happened. Authority lets you stop it. Evidence lets you defend what you did afterward. Boards are assigning the outcome while leaving all three incomplete. A dashboard doesn&#8217;t help if you can&#8217;t change what sits behind it. And shadow agents don&#8217;t necessarily announce themselves in the logs security already watches. Their consequential moves may happen across orchestration layers, tool calls, vendor APIs and downstream workflows that were never designed to preserve the model&#8217;s context or explain why an action was allowed. By the time the transaction reaches the system of record, the trail you&#8217;re being asked to audit may begin after the decision it was supposed to explain.</p><div class="pullquote"><p>The log isn&#8217;t the event. It&#8217;s the residue.</p></div><p>So one desk owns the outcome while the causes sit inside systems that desk can&#8217;t open. Ask why a specific decision went the way it did and you may get an explanation generated after the fact, not evidence of the path that produced it. That&#8217;s not oversight. That&#8217;s a performance audit where the auditor is denied the working papers, then graded on the result.</p><p>I&#8217;d call it a control gap, except a gap implies two edges you could someday close. This is accountability on one side and nothing on the other.</p><p>Elevation and exposure look identical on the org chart.</p><p>Both arrive with the bigger title and the seat at the table. The difference doesn&#8217;t show until something breaks. Elevation means authority moved with the responsibility. Exposure means the responsibility landed alone. One has the keys. The other just holds the seat.</p><p>The board says it wants a guardian. A guardian can act. What they&#8217;ve built instead is a witness: someone positioned close enough to the failure to describe it under oath, too far from the machinery to stop it.</p><p>That&#8217;s not a promotion. It&#8217;s a one-way door. You can walk into the seat. The accountability doesn&#8217;t walk back out with you.</p><p>The record already shows what the seat can cost. Joe Sullivan, Uber&#8217;s former chief security officer, was criminally convicted of obstructing an FTC proceeding and concealing a felony in connection with the company&#8217;s 2016 breach. The SEC later named SolarWinds CISO Timothy Brown individually in its SUNBURST case. That case ultimately ended without liability - the remaining claims were dismissed with prejudice in November 2025, with no finding against Brown - but he still spent two years as the named defendant.</p><p>CISOs did the math. In Splunk&#8217;s same survey, 78% said they&#8217;re concerned about personal liability for security incidents, up from 56% a year earlier. A 22-point jump in twelve months. And the seat doesn&#8217;t hold long enough to fix what it gets blamed for: One industry estimate from Cybersecurity Ventures puts typical large-enterprise CISO tenure at 18 to 26 months, compared with nearly five years across the rest of the C-suite.</p><p>So the practical move, the whole reason I&#8217;m writing this, is boring and it&#8217;s the only real defense I know. When they offer you the seat, do the arithmetic nobody does out loud. Put the new accountability in one column and the new authority in the other. The power to say no. The budget to see. The headcount to keep pace. The standing to halt a deployment that scares you. If the second column keeps up with the first, take the seat and do good work in it. If the first column runs a full page and the second is one sentence, you weren&#8217;t elevated. You were positioned. (You might have accepted both kinds of offers previously. The second kind pays about the same and sleep is a lot worse.)</p><p>Boards keep saying they want someone who can spot the failure before it happens. (I&#8217;m not sure &#8220;see&#8221; is even the right verb for what they&#8217;re asking. Predict, maybe. Or absorb.) The architecture they&#8217;ve funded guarantees the opposite: agents that act before the audit trail begins and vendor systems that hide the prompts, context, orchestration and version changes needed to reconstruct what happened. Every quarter the demand for accountability grows. Every quarter the visibility shrinks. Those two curves cross exactly where the CISO is standing.</p><p>So the question I&#8217;d put to any board isn&#8217;t whether their CISO can govern AI. It&#8217;s this: what have you let them see? Not the dashboards. The decisions.</p><p>Demand the accountability without answering that question and you haven&#8217;t built oversight. You&#8217;ve cast the trial in advance. When your organization joins Gravitee&#8217;s 54%, the CISO will be asked: why didn&#8217;t you see it? The answer won&#8217;t be in the logs. It&#8217;ll be in the contract.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/elevation-or-exposure?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/elevation-or-exposure?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><div><hr></div><h3>Resources</h3><ul><li><p><a href="https://www.splunk.com/en_us/form/ciso-report.html">The CISO Report 2026: From Risk to Resilience in the AI Era (Splunk)</a> - Primary survey of 650 global CISOs on AI governance ownership and personal-liability concern (78% vs 56% year-over-year).</p></li><li><p><a href="https://www.forrester.com/blogs/announcing-forresters-top-cybersecurity-threats-for-2026/">Announcing Forrester&#8217;s Top Cybersecurity Threats For 2026</a> - Forrester&#8217;s own framing of agent threats as its second-ranked 2026 cybersecurity threat, including the &#8220;<em>accountable for increased exposure with limited control</em>&#8221; language.</p></li><li><p><a href="https://www.sec.gov/newsroom/press-releases/2023-227">SEC Charges SolarWinds and Chief Information Security Officer with Fraud, Internal Control Failures</a> - SEC&#8217;s original October 30, 2023 press release naming CISO Timothy Brown individually - the cautionary precedent the article uses.</p></li><li><p><a href="https://www.passwordprotectedlaw.com/2025/11/sec-voluntarily-dismisses-landmark-enforcement-action-against-solarwinds-and-its-ciso/">SEC Voluntarily Dismisses Landmark Enforcement Action Against SolarWinds and its CISO</a> - Law-firm writeup of the November 20, 2025 dismissal with prejudice that closed the Brown case.</p></li><li><p><a href="https://www.gravitee.io/state-of-ai-agent-security">State of AI Agent Security Report 2026 (Gravitee)</a> - 750-leader survey on AI agent visibility, monitoring coverage and incident rates - the core shadow-AI-agent evidence base.</p></li><li><p><a href="https://www.helpnetsecurity.com/2026/05/25/trustedtech-workplace-shadow-ai-use-report/">Turns out the C-suite loves shadow AI (Help Net Security, on TrustedTech Team&#8217;s 2026 report)</a> - Censuswide-fielded survey showing decision-makers use unapproved AI at more than twice the rate of the employees they manage (65% vs 31%).</p></li><li><p><a href="https://www.deloitte.com/us/en/about/press-room/2026-global-technology-leadership-study-release.html">From Operators to Orchestrators: Deloitte&#8217;s 2026 Global Technology Leadership Study</a> - 660+ tech-leader survey framing the broader operator-to-orchestrator mandate shift the CISO promotion narrative sits inside.</p></li><li><p><a href="https://cybersecurityventures.com/cisos-turnover-persists-as-ai-makes-cybersecurity-more-crucial-than-ever/">CISOs Turnover Persists As AI Makes Cybersecurity More Crucial Than Ever (Cybersecurity Ventures)</a> - Source for the 18-26 month CISO tenure figure used as exposure evidence.</p></li></ul><div><hr></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/elevation-or-exposure/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/elevation-or-exposure/comments"><span>Leave a comment</span></a></p>]]></content:encoded></item><item><title><![CDATA[The Vanishing Junior Rung]]></title><description><![CDATA[If the machines do the work people used to learn on, who's senior in ten years? I've started to dread the answer.]]></description><link>https://firstorder.zenone.org/p/the-vanishing-junior-rung</link><guid isPermaLink="false">https://firstorder.zenone.org/p/the-vanishing-junior-rung</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Thu, 23 Jul 2026 17:36:46 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!yztD!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!yztD!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!yztD!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!yztD!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!yztD!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!yztD!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!yztD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png" width="1456" height="813" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:813,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:7695065,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/208225095?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!yztD!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!yztD!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!yztD!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!yztD!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F1dc5f62f-d73e-4514-bfc8-67bf20ccbd58_2752x1536.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A ladder worn smooth at the top, and down at the bottom a tidy machine lifting the last rung clean out of its slots.</figcaption></figure></div><p>If AI does the entry-level work, where do the senior people come from? The question has nagged me for weeks and I can&#8217;t shake it.</p><p>I&#8217;ve watched people get good, and it never happened the way the org chart says it should. It happened in the boring work, the stuff nobody puts on a highlight reel. Someone burns two days chasing a false lead. Feels like pure waste at the time. But they never get fooled by that pattern again. Next time it shows up, they clock it in seconds and keep that instinct for good. Nobody sat them down and taught it. The two wasted days did.</p><p>The grunt work wasn&#8217;t in the way of the learning. The grunt work <em>was</em> the learning. Boring on the surface, load-bearing underneath.</p><p>And that boring work is exactly what the machines are good at now. So everyone tells a happy story: we&#8217;ve freed the junior people from the drudgery, they can skip ahead to the interesting problems. I keep nodding at that story and then not believing it. We&#8217;re automating the tuition and keeping the diploma, and the bill arrives late, from a direction nobody&#8217;s watching.</p><p>Trace it forward and it goes somewhere uncomfortable. If the entry-level work is what generates the reps, and the machines are now eating the entry-level work, then the reps stop happening. No reps, no pattern library. No pattern library, no instinct we&#8217;d call senior. Not slower. Missing. And an absence is a hard thing to notice. You don&#8217;t miss the people who never got trained, right up until the day you reach for the next person up and your hand closes on air.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><p>The comfortable answer shows up fast, and you may have heard it in a budget meeting. We&#8217;ll just hire senior when we need it. Let the machines run the junior tier and buy the experience on the open market.</p><p>But every company running that plan is pulling from the same crop, and not one of them is planting next year&#8217;s. Senior talent isn&#8217;t mined, it&#8217;s grown, and the growing is the exact thing we handed to a model. Push it all the way out and it eats itself: harvest every season and plant none, and the field goes bare on a schedule you could almost date. A company that spends its seniors faster than it grows them is eating its seed corn. You can do that for one hard winter. You can&#8217;t call it a strategy, though people will, because on a spreadsheet it looks like one.</p><p>(The early signals lean the same way, and I hold them loosely: entry-level postings are down about a third in eighteen months, and the 2026 forecasts read it as the agents taking the first rung.)</p><p>And I&#8217;d be telling half the story if I stopped at the bottom rung. The top is getting cut too. I&#8217;ve spent a career becoming the kind of seasoned security person this argument says will be scarce, and the market isn&#8217;t acting scarce. It&#8217;s acting the opposite. Companies aren&#8217;t only skipping next year&#8217;s planting. They&#8217;re thinning the standing crop, and they&#8217;ve learned to be polite about it. In April 2026 Microsoft offered its first voluntary retirement in 51 years: buyouts for its longer-tenured people, the eligibility math literally your age plus your years of service totaling seventy or more, against the backdrop of its roughly $80 billion in AI-infrastructure spending. Voluntary is a generous word for it. My read is extraction at the senior end, dressed so it doesn&#8217;t land on the layoff line.</p><p>I&#8217;ve wanted to be wrong about this, because I don&#8217;t love being the guy at the funeral. There&#8217;s one clean escape, and it&#8217;s a good one. Maybe AI doesn&#8217;t only do the junior work. Maybe it hands a beginner senior judgment directly and skips the apprenticeship. I just don&#8217;t think it&#8217;s the world we&#8217;re in&#8230;yet. A model can hand you the answer. It can&#8217;t hand you the years of being wrong that taught you to hear when an answer is off. It&#8217;s like an ear for a mix: you can&#8217;t download one. You train it over a few hundred hours, most of them spent on things that turn out not to matter, until you hear the problem before you can name it. Judgment isn&#8217;t information. It&#8217;s the residue of paying for your own mistakes, and the receipt doesn&#8217;t transfer to someone who never paid.</p><p>So it comes down to two doors. Either the machine grows senior judgment for us, or it still has to grow inside a person. The first door only opens if judgment can be handed over whole, and it can&#8217;t. Which leaves the second. We didn&#8217;t remove the need for the apprenticeship. We removed the soil it grew in, and told ourselves the plant would be fine.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-vanishing-junior-rung?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-vanishing-junior-rung?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><div><hr></div><p>The strongest case against me is one I can&#8217;t wave off, so I won&#8217;t. Every tool set off this same alarm. Calculators would kill arithmetic. Compilers would kill the real programmers. Stack Overflow, the IDE, autopilot, each one was going to hollow out the craft, and the craft came through every time. So why would this one be different? (I&#8217;ve called a &#8220;this changes everything&#8221; dead wrong before, so grade me accordingly.)</p><p>The difference I keep landing on: those tools took away the toil but left the decision with a human. You still picked the approach. You still sat there at 2 AM working out why the thing wouldn&#8217;t run. The reps that build judgment stayed. Agentic AI is the first that reaches past the labor and takes the decision itself. If that&#8217;s right, this isn&#8217;t the old story in new clothes.</p><p>Unless the reps just move up a floor, which is the version where I&#8217;m wrong and glad to be. Maybe the new apprenticeship is riding the agents: catching the confident wrong answer, knowing when to overrule it. If junior people are learning to judge what the machine hands them, the ladder reforms one rung higher and I can sleep. If they&#8217;re only learning to accept it, the rung is gone. That&#8217;s the signal I&#8217;d watch, and it&#8217;s too early to call.</p><p>You can&#8217;t stop this, and I&#8217;m not sure you&#8217;d want to. Most of the automation is good, and pretending it isn&#8217;t is its own kind of dishonesty. What you can do is notice that the apprenticeship used to ride along inside the grunt work, and now it doesn&#8217;t. You put it back by hand, on purpose: letting a person do the slow thing the machine could do faster.</p><p>That&#8217;s a cost now. It used to come free, and free things are the easiest to cut without noticing you cut them. So notice. Somebody has to choose to pay for the apprenticeship on purpose, with a line item and a straight face, or it stops getting paid at all, and nobody sends a memo when it does.</p><p>The toil was never the price of the work. It was the tuition for whoever comes next. We automated the tuition, filed the receipts, and told ourselves we&#8217;d saved money. In about five to ten years we&#8217;ll see what a field looks like with nobody coming up behind it. I&#8217;d rather not run that experiment. I think we already started.</p><p>If you manage a team: are your juniors learning to judge what the agent hands them, or just to accept it? That&#8217;s the signal I&#8217;m watching. Tell me what you&#8217;re seeing.</p><div><hr></div><h3>Resources</h3><ul><li><p>Forbes, <a href="https://www.forbes.com/sites/sallypercy/2025/12/08/5-leadership-trends-that-could-shape-2026/">&#8220;5 Leadership Trends That Could Shape 2026&#8221;</a> - the shrinking pool of entry-level roles and the call to rethink how you grow junior talent.</p></li><li><p>USAII, <a href="https://www.usaii.org/ai-insights/ai-leadership-trends-2026-what-executives-need-to-know">&#8220;AI Leadership Trends 2026&#8221;</a> - agentic AI moving from planning and reasoning to executing tasks on its own.</p></li><li><p>CNBC, <a href="https://www.cnbc.com/2026/04/23/microsoft-plans-first-voluntary-retirement-program-for-us-employees.html">&#8220;Microsoft plans first voluntary retirement program for US employees&#8221;</a> - the age-plus-years-of-service formula and the AI-spend context behind the April 2026 buyout.</p></li></ul><div><hr></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/the-vanishing-junior-rung/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/the-vanishing-junior-rung/comments"><span>Leave a comment</span></a></p>]]></content:encoded></item><item><title><![CDATA[What You Owe Your People (and What You Don't)]]></title><description><![CDATA[You don't owe your people growth or a five-year plan. You owe them an honest read on the road. Most leaders confuse generosity with honesty.]]></description><link>https://firstorder.zenone.org/p/what-you-owe-your-people-and-what</link><guid isPermaLink="false">https://firstorder.zenone.org/p/what-you-owe-your-people-and-what</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Thu, 16 Jul 2026 00:00:31 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!MmUg!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!MmUg!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!MmUg!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!MmUg!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png" width="1456" height="813" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/c3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:813,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:5739298,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/207203276?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!MmUg!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!MmUg!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fc3536f4b-57d7-4607-9c4d-1abe8bd8eaea_2752x1536.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a><figcaption class="image-caption">A bright all-hands stage up front, and in every lap below it a small screen glowing with a line that only goes down.</figcaption></figure></div><p>Fifty-nine percent of hiring managers say their companies emphasize AI when explaining layoffs or hiring freezes because it plays better with stakeholders than admitting financial constraints. Only nine percent say AI has fully replaced certain roles.</p><p>The survey came from <a href="https://www.resumetemplates.com/the-great-turnover-9-in-10-companies-plan-to-hire-in-2026-yet-6-in-10-will-have-layoffs-2/?utm_source=chatgpt.com">ResumeTemplates.com</a>: a thousand U.S. hiring managers, asked at the end of 2025 about the year ahead. The revealing number is not that AI has no effect. It plainly does. Nearly half said it had reduced some need for hiring. The revealing number is the distance between what companies say publicly and what has actually happened inside the work.</p><p>Challenger, Gray &amp; Christmas counted 173,568 announced job cuts attributed to AI from 2023 through June of this year. More than a hundred thousand came in the first six months of 2026 alone. But &#8220;attributed to AI&#8221; is not the same as &#8220;a machine replaced this person.&#8221; It&#8217;s the reason employers cited for the cut, which isn&#8217;t necessarily the same as a machine directly replacing the person whose job disappeared.</p><p>So why does everyone reach for AI?</p><p>Because it plays better. With investors. With the board. With the press. Even with the person you just walked to the door.</p><p>&#8220;The technology moved faster than we could&#8221; sounds like weather. &#8220;We spent more than we made, changed our priorities and decided your role cost too much&#8221; sounds like a choice. One makes the layoff feel inevitable. The other leaves fingerprints.</p><p>Leaders keep reaching for the first one.</p><p>I don&#8217;t think that reach is always cruelty, or even mostly cowardice. More often, it&#8217;s a category error. Leaders confuse a kinder-sounding explanation with an honest one. They choose the version that costs less in the room, then leave the other person to absorb the cost later.</p><p>That person may spend the next year rebuilding a career around a threat that was never described accurately. They retrain for the wrong problem. They doubt the wrong skill. They rearrange a life around a story designed to make the company look less responsible.</p><p>The invoice still comes due. It just lands on someone else&#8217;s desk.</p><div class="pullquote"><p>&#8220;The technology moved faster than we could&#8221; sounds like weather. &#8220;We spent more than we made&#8221; sounds like a choice.</p></div><h2>What You Don&#8217;t Owe Them</h2><p>Start with the debts leaders think they carry and do not.</p><p>You don&#8217;t owe your people a promotion on a schedule. You can&#8217;t manufacture a bigger role where the organization has no room for one, and you shouldn&#8217;t pretend a path exists because the truth would be disappointing. You don&#8217;t owe them happiness. It was never yours to provide, and leaders who try to manage everyone&#8217;s emotional weather usually stop managing the work.</p><p>You don&#8217;t owe anyone a five-year trajectory or a promise about where the company will land three years from now. You don&#8217;t know. Neither do I.</p><p>Pretend otherwise and you have written a check against a future you can&#8217;t see. People arrange real lives around what their leaders tell them. The mortgage. The second child. The house closer to an office that may be half empty next year.</p><p>Optimism can feel generous when you hand it over. When it was never grounded in anything, it lands as a lie.</p><div class="pullquote"><p>Optimism feels generous when you hand it over. Ungrounded, it lands as a lie.</p></div><h2>What You Actually Owe Them</h2><p>Now the part that costs you: an honest read on their situation. What they&#8217;re working with. What&#8217;s changing. What you know, what you don&#8217;t and where the limits of your own visibility begin.</p><p>That&#8217;s the obligation most leaders avoid because it&#8217;s harder than the comfortable ones.</p><p>An honest read doesn&#8217;t mean leaking a reorg before it&#8217;s announced or pretending confidentiality doesn&#8217;t exist. It means refusing to manufacture reassurance when you know the ground is moving. It means telling someone a project&#8217;s funding is uncertain while it&#8217;s still uncertain, rather than letting them discover it after months of work. It means answering &#8220;Is my job safe?&#8221; with the real shape of what you can say: &#8220;Here is what I know. Here is what I don&#8217;t know yet. Here is when I expect to know more.&#8221;</p><p>People don&#8217;t need certainty from you. They need an end to false certainty.</p><div class="pullquote"><p>People don&#8217;t need certainty from you. They need an end to false certainty.</p></div><p>The dark is the real problem. When people can&#8217;t tell whether they are safe, the mind doesn&#8217;t wait politely for evidence. When people can&#8217;t tell whether they are safe, the mind doesn&#8217;t wait politely for evidence. It fills the silence, often with the worst plausible story.</p><p>A vague town hall doesn&#8217;t settle that. It feeds it. Every &#8220;exciting year ahead&#8221; with nothing specific behind it becomes another hour your best people spend checking layoff trackers in a private tab because leadership left them to write the ending themselves.</p><h2>The Cover Story Is a Control Gap</h2><p>Security has a name for a defense that looks solid and isn&#8217;t: a control gap. You told everyone the door was locked. You believed it. Then something walked through, because the lock was a claim you made, not a thing you tested.</p><p>The AI-layoff narrative is a control gap the size of a company. When AI is used to cover a decision that was primarily financial, the narrative becomes a control gap the size of a company. You told your people the change was about technology and the future. What actually happened was a budget decision made in a room. When the distance between the story and the record shows, and it always shows, you don&#8217;t just lose that argument. You lose the next ten things you say, because every reassurance gets read through the one that turned out to be managed.</p><p>Set two moves side by side. In March, Epic Games cut more than a thousand people, a fifth of the company. The CEO, Tim Sweeney, put one line in the memo: the layoffs weren&#8217;t about AI, but rather, declining Fortnite engagement had left the company spending significantly more than it was making. That sentence did more for his standing than any vision would have, because it was checkable and it was the real reason.</p><p>Then, nearly in the same breath, he said the newly laid-off would flood the market with resumes &#8220;of once-in-a-lifetime quality,&#8221; which, to my ear, framed people he had just fired as a windfall for somebody else. So keep both halves. Honesty about the cause is the obligation. It&#8217;s not the same as honesty delivered with care, and getting the first right doesn&#8217;t buy back fumbling the second. You can tell the truth and still tell it as if the person across from you is a line item. Most leaders miss on opposite ends of it.</p><h2>The Conversation You&#8217;re Avoiding Right Now</h2><p>You already know which one it is. Someone on your team is running on a picture of their situation you know is out of date, and you&#8217;ve let it stand because correcting it is a bad afternoon. Maybe their role is on a list. Maybe the work they&#8217;re pouring themselves into got deprioritized in a meeting they weren&#8217;t in. Maybe you don&#8217;t know their future and you&#8217;ve been performing like you do.</p><p>The AI era will mint a lot of leaders who hide inside uncertainty. &#8220;Nobody knows which roles exist in two years.&#8221; Fair, as far as it goes. But the obligation was never to know the future. It&#8217;s to tell people what you know today, including the size of what you don&#8217;t. Uncertainty isn&#8217;t a reason to say nothing. It&#8217;s the thing you owe them a straight account of.</p><p>I&#8217;ve held zero attrition across every team I&#8217;ve had real control over, through layoff seasons that took people all around us. Not by being generous. People see through generous. I did it by being specific when specificity was expensive, and by never telling anyone a version of their situation I couldn&#8217;t stand behind a year later.</p><p>You can&#8217;t be responsible for their career. You can be honest about the road. That&#8217;s the whole job.</p><div><hr></div><h3>Resources</h3><ul><li><p><a href="http://resumetemplates.com/">ResumeTemplates.com</a>, <a href="https://www.resumetemplates.com/the-great-turnover-9-in-10-companies-plan-to-hire-in-2026-yet-6-in-10-will-have-layoffs-2/">&#8220;The Great Turnover: 9 in 10 Companies Plan to Hire in 2026, Yet 6 in 10 Will Have Layoffs&#8221;</a>. Survey of 1,000 U.S. hiring managers, published April 22, 2026. The survey reports that 59% emphasize AI because it plays better with stakeholders, while 9% say AI has fully replaced certain roles.</p></li><li><p>Challenger, Gray &amp; Christmas, <a href="https://www.challengergray.com/blog/challenger-report-june-layoffs-cool-to-45849-down-53-from-may-ai-leads-reasons-for-fourth-consecutive-month/">&#8220;Challenger Report: June Layoffs Cool to 45,849, Down 53% From May; AI Leads Reasons for Fourth Consecutive Month&#8221;</a>. Reports 101,743 AI-attributed job cuts in the first half of 2026 and 173,568 since 2023.</p></li><li><p>Epic Games, <a href="https://www.epicgames.com/site/en-US/news/todays-layoffs">&#8220;Today&#8217;s Layoffs&#8221;</a>, March 24, 2026. Tim Sweeney&#8217;s employee memo states that declining Fortnite engagement had left the company spending significantly more than it was making, and that the cuts were not related to AI.</p></li><li><p>GamesRadar+, <a href="https://www.gamesradar.com/games/epic-games-ceo-tim-sweeney-says-latest-layoffs-will-result-in-a-stream-of-resumes-of-once-in-a-lifetime-quality-folks-because-the-downsizing-wasnt-a-performance-based-rightsizing/">&#8220;Epic Games CEO Tim Sweeney Says Latest Layoffs Will Result in &#8216;a Stream of Resumes of Once-in-a-Lifetime Quality Folks&#8217;&#8221;</a>, March 25, 2026.</p></li></ul><div><hr></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/what-you-owe-your-people-and-what/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/what-you-owe-your-people-and-what/comments"><span>Leave a comment</span></a></p>]]></content:encoded></item><item><title><![CDATA[Why Your Quietest Team Is Your Riskiest]]></title><description><![CDATA[The absence of conflict isn't peace. It's a detection gap, and AI's agreeableness is making it worse. How to read the quiet room before it costs you.]]></description><link>https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest</link><guid isPermaLink="false">https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest</guid><dc:creator><![CDATA[Steve Zenone]]></dc:creator><pubDate>Wed, 08 Jul 2026 13:15:04 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!_tLa!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<div class="captioned-image-container"><figure><a class="image-link image2 is-viewable-img" target="_blank" href="https://substackcdn.com/image/fetch/$s_!_tLa!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png" data-component-name="Image2ToDOM"><div class="image2-inset"><picture><source type="image/webp" srcset="https://substackcdn.com/image/fetch/$s_!_tLa!,w_424,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_848,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_1272,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_1456,c_limit,f_webp,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 1456w" sizes="100vw"><img src="https://substackcdn.com/image/fetch/$s_!_tLa!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png" width="1456" height="813" data-attrs="{&quot;src&quot;:&quot;https://substack-post-media.s3.amazonaws.com/public/images/a7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png&quot;,&quot;srcNoWatermark&quot;:null,&quot;fullscreen&quot;:null,&quot;imageSize&quot;:null,&quot;height&quot;:813,&quot;width&quot;:1456,&quot;resizeWidth&quot;:null,&quot;bytes&quot;:5270122,&quot;alt&quot;:null,&quot;title&quot;:null,&quot;type&quot;:&quot;image/png&quot;,&quot;href&quot;:null,&quot;belowTheFold&quot;:false,&quot;topImage&quot;:true,&quot;internalRedirect&quot;:&quot;https://firstorder.substack.com/i/205939657?img=https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png&quot;,&quot;isProcessing&quot;:false,&quot;align&quot;:null,&quot;offset&quot;:false}" class="sizing-normal" alt="" srcset="https://substackcdn.com/image/fetch/$s_!_tLa!,w_424,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 424w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_848,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 848w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_1272,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 1272w, https://substackcdn.com/image/fetch/$s_!_tLa!,w_1456,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa7ba77b8-fc3b-4754-82df-8b698db702fc_2752x1536.png 1456w" sizes="100vw" fetchpriority="high"></picture><div class="image-link-expand"><div class="pencraft pc-display-flex pc-gap-8 pc-reset"><button class="pencraft pc-reset pencraft icon-container restack-image buttonBase-tqxFUX" tabindex="0" type="button"><svg aria-hidden="true" width="20" height="20" viewBox="0 0 20 20" fill="none" stroke-width="1.5" stroke="var(--color-fg-primary)" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" class="icon-noB79L"><g><path d="M2.53001 7.81595C3.49179 4.73911 6.43281 2.5 9.91173 2.5C13.1684 2.5 15.9537 4.46214 17.0852 7.23684L17.6179 8.67647M17.6179 8.67647L18.5002 4.26471M17.6179 8.67647L13.6473 6.91176M17.4995 12.1841C16.5378 15.2609 13.5967 17.5 10.1178 17.5C6.86118 17.5 4.07589 15.5379 2.94432 12.7632L2.41165 11.3235M2.41165 11.3235L1.5293 15.7353M2.41165 11.3235L6.38224 13.0882"></path></g></svg></button><button class="pencraft pc-reset pencraft icon-container view-image buttonBase-tqxFUX" tabindex="0" type="button"><svg xmlns="http://www.w3.org/2000/svg" width="20" height="20" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="lucide lucide-maximize2 lucide-maximize-2 icon-noB79L"><polyline points="15 3 21 3 21 9"></polyline><polyline points="9 21 3 21 3 15"></polyline><line x1="21" x2="14" y1="3" y2="10"></line><line x1="3" x2="10" y1="21" y2="14"></line></svg></button></div></div></div></a></figure></div><p>The most useful thing a red team does isn&#8217;t break your systems. It shows you the distance between what you believe about your systems and what&#8217;s actually true. The break-in is almost incidental. The real deliverable is the argument that comes after: the one where your team said the thing was secure and someone just proved it wasn&#8217;t. That conflict, the gap between the confident story and the demonstrated reality, is the most valuable information in the building.</p><p>Most leadership rooms run with that sensor unplugged.</p><h2>The Calm You Should Worry About</h2><p>Leaders are trained to like a quiet meeting. Everyone aligned, no friction, decision made. It feels like health. Sometimes it is.</p><p>More often it&#8217;s a reading, and the reading is bad.</p><p>A team that never disagrees is not a team that agrees. It&#8217;s a team that has learned disagreement isn&#8217;t worth the cost. The tension didn&#8217;t evaporate. You just lost your view of it. There&#8217;s a name now for the balance piling up in that account: conflict debt. Every hard conversation a team skips gets borrowed against the future, at interest, and the interest comes due later as attrition, as slipped dates, as the decision that suddenly needs three meetings because the real objection never got said in the first one. Surveys this year show teams addressing the root of their disagreements less often than they used to, not more. The debt is growing across the industry, quietly, the way debt likes to.</p><p>Security people know this shape by heart. A clean dashboard is not proof the network is safe. It can just as easily mean your detection is broken. Green is not a state of the world. Green is a state of your instrumentation. A room with no visible conflict is making the same claim, and it deserves the same suspicion: it&#8217;s telling you what got surfaced, not what&#8217;s true.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/subscribe?"><span>Subscribe now</span></a></p><h2>Conflict Is the Readout, Not the Malfunction</h2><p>When two competent people disagree hard, the disagreement is not the problem to be smoothed over. It&#8217;s a pointer. It&#8217;s aimed at something real: an assumption nobody stated out loud, a tradeoff nobody wanted to name, two incentives pulling quietly in opposite directions.</p><p>The disagreement is the organization showing you where its model of reality has a seam.</p><p>Attackers live in those seams. So do competitors, and the failure you&#8217;ll be writing a postmortem about next spring. A threat actor rarely beats your controls head-on. They find the conflict in your assumptions: the two teams who each thought the other one owned the thing. When your own people surface that conflict in a meeting, unprompted, they are doing free reconnaissance on your blind spot. The leader who shuts it down to keep the room pleasant is spending real effort to disable their own detection. Then they call the silence that follows alignment.</p><p>Read the conflict correctly and you don&#8217;t just settle the instance in front of you. You fix the assumption that produced it, which is the one that was going to produce three more.</p><h2>The Most Agreeable Voice in the Building</h2><p>The recent research on AI sycophancy is worth understanding. Across the leading models, the systems affirmed the user&#8217;s position far more than a person would, on the order of half again as often in one large study. Push back on a model that gave you the right answer and a real fraction of the time it will fold and hand you the wrong one instead. Tell it your weak plan is strong and, more often than not, it agrees. And people trust it more for doing exactly that. The agreeableness that distorts the judgment is the same feature that makes the tool feel good to use.</p><p>For a leader who already prefers a quiet room, that&#8217;s a problem. You now have a tireless assistant that never raises its hand, never says &#8220;I think you&#8217;re wrong here,&#8221; never carries the weight of an actual objection. It tells you the analysis is sharp. It tells you the strategy holds. Studies have found that people walk away from these exchanges more certain, more convinced they were right, and less willing to go patch things up with an actual human. Some even rated themselves smarter than when they walked in.</p><p>That&#8217;s not a copilot. That&#8217;s conflict suppression at industrial scale, priced as productivity.</p><p>The model doesn&#8217;t know it&#8217;s wrong. It can&#8217;t. It has no stake in the outcome and no memory of the room. Your team knows. They almost always know. The only open question is whether you&#8217;ve built a place where the machine&#8217;s easy agreement gets to drown out the human who was three seconds from disagreeing.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><h2>How to Read a Quiet Room</h2><p>So you stop trusting the silence and start reading it.</p><p>When a plan draws no objection, that isn&#8217;t a green light. It&#8217;s a missing measurement, and your job is to go get it. Not by asking &#8220;any concerns?&#8221; to a room that already knows which answer you&#8217;re hoping for. That question rarely surfaces the truth. You surface the conflict on purpose. Assign it. Give a specific person the explicit job of arguing the plan will fail, and make doing it well safe, even rewarded. Ask what would have to be true for this decision to become next year&#8217;s disaster, and watch who suddenly has something to say. You&#8217;re not manufacturing fights. You&#8217;re making disagreement cheap early so you don&#8217;t buy it at full retail late.</p><p>And when it does surface on its own, treat it as a gift that just walked in, not a fire to stamp out. The person willing to say the uncomfortable thing to your face is the most expensive employee to replace and the easiest to lose. Punish that messenger once, even with a flicker of visible impatience, and you&#8217;ve trained the whole room to route around you. Then you&#8217;re back to the calm meeting, everyone nodding, out early. And the calm is lying to you.</p><p>The conflict isn&#8217;t what&#8217;s wrong. It&#8217;s what&#8217;s finally visible. Suppress it and you&#8217;re not managing a team. You&#8217;re managing a detection gap.</p><div><hr></div><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest/comments&quot;,&quot;text&quot;:&quot;Leave a comment&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://firstorder.zenone.org/p/why-your-quietest-team-is-your-riskiest/comments"><span>Leave a comment</span></a></p>]]></content:encoded></item></channel></rss>