{"id":1064,"date":"2026-11-17T12:22:46","date_gmt":"2026-11-17T12:22:46","guid":{"rendered":"https:\/\/managedt.com\/blog\/ai-finds-28-zero-day-flaws-in-robotics-software\/"},"modified":"2026-11-17T12:22:47","modified_gmt":"2026-11-17T12:22:47","slug":"ai-finds-28-zero-day-flaws-in-robotics-software","status":"publish","type":"post","link":"https:\/\/managedt.com\/blog\/ai-finds-28-zero-day-flaws-in-robotics-software\/","title":{"rendered":"AI Finds 28 Zero-Day Flaws in Robotics Software"},"content":{"rendered":"<div class=\"post-updated\"><em>Updated November 17, 2026 at 6:57 AM ET<\/em><\/div>\n<div class=\"post-quick-answer\" style=\"background:rgba(128,128,128,.08);border-left:4px solid rgba(128,128,128,.6);padding:14px 18px;margin:0 0 24px\">\n<p class=\"post-quick-answer-label\"><strong>Quick Answer<\/strong><\/p>\n<p>AI-assisted research identified 28 zero-day vulnerabilities in less than two days across software and tools used to develop, train, simulate and operate robots. VicOne, the cybersecurity company behind the research, said the count may change after vendor and vulnerability-authority review, and some findings do not yet have CVE tracking numbers.<\/p>\n<\/div>\n<p><!--\/post-quick-answer--><\/p>\n<p>The findings affect projects developed or maintained by Google, NVIDIA and Meta, among others. They show how quickly AI can uncover previously unknown weaknesses in software connected to robotics, while leaving teams to determine whether their systems use the affected tools and how the flaws could affect operations.<\/p>\n<nav class=\"post-contents\" style=\"border:1px solid rgba(128,128,128,.35);border-radius:6px;padding:14px 18px;margin:0 0 28px\">\n<p><strong>In This Article<\/strong><\/p>\n<ul>\n<li><a href=\"#what-happened\">What Happened<\/a><\/li>\n<li><a href=\"#anthropic-s-2026-zero-day-discovery\">Anthropic&#8217;s 2026 Zero-Day Discovery<\/a><\/li>\n<li><a href=\"#are-you-affected\">Are You Affected?<\/a><\/li>\n<li><a href=\"#what-to-do-now\">What to Do Now<\/a><\/li>\n<li><a href=\"#where-the-findings-sit-in-the-robot-lifecycle\">Where the Findings Sit in the Robot Lifecycle<\/a><\/li>\n<li><a href=\"#what-the-evidence-shows\">What the Evidence Shows<\/a><\/li>\n<li><a href=\"#why-ai-changes-the-response-window\">Why AI Changes the Response Window<\/a><\/li>\n<li><a href=\"#key-considerations\">Key Considerations<\/a><\/li>\n<li><a href=\"#faq\">FAQ<\/a><\/li>\n<li><a href=\"#related-coverage\">Related Coverage<\/a><\/li>\n<\/ul>\n<\/nav>\n<h2 id=\"what-happened\" data-toc=\"1\">What Happened<\/h2>\n<p>VicOne said its AI-assisted research identified an initial set of 28 zero-day findings in less than two days. The company said all 28 were classified as zero-day vulnerabilities after initial validation, though the total may change following review by vendors and vulnerability authorities.<\/p>\n<p>The research covered software and tools in the Physical AI ecosystem, including projects developed or maintained by Google, NVIDIA and Meta. <a href=\"https:\/\/spectrum.ieee.org\/zero-day-robotics-vulnerabilities-vicone-security\" target=\"_blank\" rel=\"noopener\">IEEE Spectrum reported<\/a> that some findings do not yet have CVE tracking numbers.<\/p>\n<h2 id=\"anthropic-s-2026-zero-day-discovery\" data-toc=\"1\">Anthropic&#8217;s 2026 Zero-Day Discovery<\/h2>\n<h2 id=\"are-you-affected\" data-toc=\"1\">Are You Affected?<\/h2>\n<p>Teams need to determine whether they use the affected software and what an attacker could access or change. <a href=\"https:\/\/spectrum.ieee.org\/zero-day-robotics-vulnerabilities-vicone-security\" target=\"_blank\" rel=\"noopener\">According to IEEE Spectrum<\/a>, VicOne\u2019s findings include projects developed or maintained by Google, NVIDIA and Meta, among others.<\/p>\n<p>Businesses using robotics software should check whether their development, training or simulation tools, and their operational monitoring or remote-operation systems, use the projects identified by VicOne.<\/p>\n<p>The 28 findings were vulnerabilities identified through research, not confirmed attacks. Their effects depend on an attacker\u2019s access, the system\u2019s design and safeguards that prevent changes from reaching physical robot behavior.<\/p>\n<h2 id=\"what-to-do-now\" data-toc=\"1\">What to Do Now<\/h2>\n<ol>\n<li>\n<p>Inventory the robotics development, training, simulation, validation and runtime software used by your organization.<\/p>\n<\/li>\n<li>\n<p>Compare that software with the projects and tools identified by VicOne, including projects maintained by Google, NVIDIA and Meta.<\/p>\n<\/li>\n<li>\n<p>Review whether each matching component could affect training data, models, test results, motion planning, operator monitoring or instructions sent to robots.<\/p>\n<\/li>\n<li>\n<p>Test whether remote-stop and other protective functions remain effective when supporting software is compromised.<\/p>\n<\/li>\n<li>\n<p>Review any vendor fixes for the identified findings and determine whether they are safe to deploy before applying them in operational systems.<\/p>\n<\/li>\n<\/ol>\n<h2 id=\"where-the-findings-sit-in-the-robot-lifecycle\" data-toc=\"1\">Where the Findings Sit in the Robot Lifecycle<\/h2>\n<p>VicOne\u2019s research covers three stages of the robotics lifecycle: development and training, simulation and validation, and deployment and runtime. In this set, 75% of findings involved development, training, simulation or validation tools, placing most of the discovered risks before a robot enters operation.<\/p>\n<figure class=\"post-section-image\"><img decoding=\"async\" data-src=\"https:\/\/managedt.com\/blog\/wp-content\/smush-avif\/2026\/11\/zero-day-robotics-vulnerabilities-ai-finds-28-flaws-software-2.png.avif\" alt=\"Robotic arm with gripper surrounded by glowing wireframe mesh in workshop\" width=\"1536\" height=\"1024\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1536px; --smush-placeholder-aspect-ratio: 1536\/1024;\" data-smush-avif-fallback=\"{&quot;data-src&quot;:&quot;https:\\\/\\\/managedt.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/zero-day-robotics-vulnerabilities-ai-finds-28-flaws-software-2.png&quot;}\"><\/figure>\n<h3>How Could Development and Training Tools Be Affected?<\/h3>\n<p>A manipulated model might behave normally during routine testing but respond incorrectly to a particular object, scene or command after deployment.<\/p>\n<h3>How Could Simulation and Validation Results Be Misleading?<\/h3>\n<p>A compromised scene or robot model could give an attacker access to a simulation environment. With enough control, the attacker could alter simulated sensor readings, physical conditions or test results, allowing a robot to pass tests that should have exposed unsafe behavior.<\/p>\n<h3>How Could Deployment and Runtime Systems Be Affected?<\/h3>\n<p>Depending on the system and its safeguards, the effects could include altered robot movement or a loss of operator visibility and control.<\/p>\n<h2 id=\"what-the-evidence-shows\" data-toc=\"1\">What the Evidence Shows<\/h2>\n<p>VicOne reported an initial set of 28 zero-day findings across software and tools supporting the Physical AI ecosystem. The company identified them in less than two days with AI-assisted research and classified all 28 as zero-day vulnerabilities after initial validation.<\/p>\n<p>The company said the count may change after vendor and vulnerability-authority review, and some findings do not yet have CVE tracking numbers. Seventy-five percent of the findings involved development, training, simulation or validation tools.<\/p>\n<p>One case showed that a protective function could be weakened. It did not show that every finding had the same effect.<\/p>\n<p><a href=\"https:\/\/labs.cloudsecurityalliance.org\/research\/csa-research-note-ai-generated-zero-day-exploits-adversarial\/\" target=\"_blank\" rel=\"noopener\">Cloud Security Alliance\u2019s AI Safety Initiative reported<\/a> that Google\u2019s Threat Intelligence Group had confirmed what it described as the first in-the-wild zero-day exploit it believed was discovered and developed with an AI model.<\/p>\n<p>The same research note said a criminal threat actor planned to weaponize the vulnerability in a mass exploitation campaign, which Google\u2019s Threat Intelligence Group disrupted through detection and coordinated disclosure before exploitation began at scale. It also reported that Google\u2019s Threat Intelligence Group tracked 90 zero-days exploited in the wild during 2025, including 43 targeting enterprise technologies.<\/p>\n<h2 id=\"why-ai-changes-the-response-window\" data-toc=\"1\">Why AI Changes the Response Window<\/h2>\n<p>Researchers find a vulnerability by reading code, tracing how systems talk to each other and testing whether a suspected weakness can be exploited. AI can help them review more software in less time. Attackers could use similar tools.<\/p>\n<figure class=\"post-section-image\"><img decoding=\"async\" data-src=\"https:\/\/managedt.com\/blog\/wp-content\/smush-avif\/2026\/11\/zero-day-robotics-vulnerabilities-ai-finds-28-flaws-software-3.png.avif\" alt=\"Robotic arm positioned beside a cracked glass panel\" width=\"1536\" height=\"1024\" src=\"data:image\/svg+xml;base64,PHN2ZyB3aWR0aD0iMSIgaGVpZ2h0PSIxIiB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciPjwvc3ZnPg==\" class=\"lazyload\" style=\"--smush-placeholder-width: 1536px; --smush-placeholder-aspect-ratio: 1536\/1024;\" data-smush-avif-fallback=\"{&quot;data-src&quot;:&quot;https:\\\/\\\/managedt.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/zero-day-robotics-vulnerabilities-ai-finds-28-flaws-software-3.png&quot;}\"><\/figure>\n<p>A zero-day has no available fix at the time it is discovered, so faster discovery can leave a team with less time to assess exposure and respond. VicOne\u2019s findings therefore place two demands on robotics teams: responding as weaknesses are identified faster and determining how each finding could affect robot safety and operations.<\/p>\n<p>Before applying a published patch to an operational system, check whether it is safe to deploy in that system.<\/p>\n<h2 id=\"key-considerations\" data-toc=\"1\">Key Considerations<\/h2>\n<ul>\n<li>\n<p>The total of 28 findings may change after vendor and vulnerability-authority review.<\/p>\n<\/li>\n<li>\n<p>Some findings do not yet have CVE tracking numbers.<\/p>\n<\/li>\n<li>\n<p>The reported effects are potential consequences whose actual impact depends on attacker access, system design and existing safeguards.<\/p>\n<\/li>\n<li>\n<p>Affected software can sit in development, training, simulation, validation, deployment and runtime environments.<\/p>\n<\/li>\n<li>\n<p>Most findings in this research set, 75%, affected tools used before robots were deployed.<\/p>\n<\/li>\n<\/ul>\n<h2 id=\"faq\" data-toc=\"1\">FAQ<\/h2>\n<h3>How Many Zero-Day Vulnerabilities Did AI-assisted Research Find?<\/h3>\n<p>VicOne identified an initial set of 28 zero-day findings in less than two days across software and tools supporting the Physical AI ecosystem.<\/p>\n<h3>Which Robotics Software Was Affected?<\/h3>\n<p>VicOne\u2019s findings covered projects developed or maintained by Google, NVIDIA and Meta, among others.<\/p>\n<h3>Were the 28 Findings Confirmed Attacks?<\/h3>\n<p>No. The findings were vulnerabilities identified through research. The source described potential effects that depend on attacker access, system design and existing safeguards, rather than confirmed attacks against every affected system.<\/p>\n<h3>Can AI-assisted Discovery Make Zero-Day Response More Urgent?<\/h3>\n<p>AI can help researchers review more software in less time. VicOne found 28 zero-day vulnerabilities in less than two days. Attackers could use similar tools. A zero-day has no available fix when it is discovered, so faster discovery may leave teams with less time to check exposure and respond.<\/p>\n<h3>What Should Robotics Teams Check After a Patch Is Published?<\/h3>\n<p>Teams should determine whether the fix is safe to deploy in their system. They should also test whether safeguards such as remote-stop functions remain effective when supporting software is compromised.<\/p>\n<h2 id=\"related-coverage\" data-toc=\"1\">Related Coverage<\/h2>\n<ul>\n<li><a href=\"https:\/\/managedt.com\/blog\/check-point-management-server-zero-day-targeted-attacks\/\">Check Point Warns of Management Server Zero-Day Exploited in Targeted Attacks<\/a> (September 23, 2026): Check Point management server administrators can now close a critical zero-day path traversal flaw that attackers exploited in targeted attacks, while Spark firewall customers need to verify that the September 9 VPN fix is installed.<\/li>\n<\/ul>\n<div class=\"source-disclaimer\"><em>This article reports on publicly available information from the sources listed below. We have no reason to believe it is inaccurate, but we have not independently verified every claim, and details may change as the story develops.<\/em><\/div>\n<section class=\"post-sources\">\n<h2>Sources<\/h2>\n<ul>\n<li><a href=\"https:\/\/www.zdnet.com\/innovation\/ai-finds-zero-day-security-flaws-for-cheap-hackers\/\" target=\"_blank\" rel=\"noopener nofollow\">zdnet.com<\/a><\/li>\n<li><a href=\"https:\/\/spectrum.ieee.org\/zero-day-robotics-vulnerabilities-vicone-security\" target=\"_blank\" rel=\"noopener nofollow\">spectrum.ieee.org<\/a><\/li>\n<li><a href=\"https:\/\/labs.cloudsecurityalliance.org\/research\/csa-research-note-ai-generated-zero-day-exploits-adversarial\/\" target=\"_blank\" rel=\"noopener nofollow\">labs.cloudsecurityalliance.org<\/a><\/li>\n<li><a href=\"https:\/\/www.anthropic.com\/research\/zero-days\" target=\"_blank\" rel=\"noopener nofollow\">anthropic.com<\/a><\/li>\n<\/ul>\n<\/section>\n<p><script type=\"application\/ld+json\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"FAQPage\",\"mainEntity\":[{\"@type\":\"Question\",\"name\":\"How Many Zero-Day Vulnerabilities Did AI-assisted Research Find?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"VicOne identified an initial set of 28 zero-day findings in less than two days across software and tools supporting the Physical AI ecosystem.\"}},{\"@type\":\"Question\",\"name\":\"Which Robotics Software Was Affected?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"VicOne\u2019s findings covered projects developed or maintained by Google, NVIDIA and Meta, among others.\"}},{\"@type\":\"Question\",\"name\":\"Were the 28 Findings Confirmed Attacks?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"No. The findings were vulnerabilities identified through research. The source described potential effects that depend on attacker access, system design and existing safeguards, rather than confirmed attacks against every affected system.\"}},{\"@type\":\"Question\",\"name\":\"Can AI-assisted Discovery Make Zero-Day Response More Urgent?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"AI can help researchers review more software in less time. VicOne found 28 zero-day vulnerabilities in less than two days. Attackers could use similar tools. A zero-day has no available fix when it is discovered, so faster discovery may leave teams with less time to check exposure and respond.\"}},{\"@type\":\"Question\",\"name\":\"What Should Robotics Teams Check After a Patch Is Published?\",\"acceptedAnswer\":{\"@type\":\"Answer\",\"text\":\"Teams should determine whether the fix is safe to deploy in their system. They should also test whether safeguards such as remote-stop functions remain effective when supporting software is compromised.\"}}]}]}<\/script><\/p>\n","protected":false},"excerpt":{"rendered":"<p>VicOne identified 28 previously unknown vulnerabilities in software used to develop, train, simulate and operate robots. The findings may change after vendor<\/p>\n","protected":false},"author":3,"featured_media":1061,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-1064","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/posts\/1064","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/comments?post=1064"}],"version-history":[{"count":1,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/posts\/1064\/revisions"}],"predecessor-version":[{"id":1065,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/posts\/1064\/revisions\/1065"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/media\/1061"}],"wp:attachment":[{"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/media?parent=1064"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/categories?post=1064"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/managedt.com\/blog\/wp-json\/wp\/v2\/tags?post=1064"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}