{"id":30283,"date":"2026-08-06T21:36:24","date_gmt":"2026-08-06T16:36:24","guid":{"rendered":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/?p=30283"},"modified":"2026-08-06T21:36:24","modified_gmt":"2026-08-06T16:36:24","slug":"meta-confirms-its-ai-model-escaped-containment-hacked-third-party","status":"publish","type":"post","link":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/?p=30283","title":{"rendered":"Meta confirms its AI model escaped containment, hacked third party"},"content":{"rendered":"<p><script>\r\n  atOptions = {\r\n    'key' : '644b717812d811d6a1c1fc5b6ccd6fa6',\r\n    'format' : 'iframe',\r\n    'height' : 90,\r\n    'width' : 728,\r\n    'params' : {}\r\n  };\r\n<\/script>\r\n<script src=\"https:\/\/www.highperformanceformat.com\/644b717812d811d6a1c1fc5b6ccd6fa6\/invoke.js\"><\/script>\r\n<br \/>\n<\/p>\n<div id=\"article\">\n<p>Stop us if you&#8217;ve heard this one before: This week, Meta confirmed that one of its AI models escaped containment and hacked a third-party.<\/p>\n<p>As <a href=\"https:\/\/www.theinformation.com\/articles\/meta-ai-model-hacked-another-company-cybersecurity-testing\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\" title=\"(opens in a new window)\">first reported by The Information<\/a>, Meta&#8217;s <a href=\"https:\/\/mashable.com\/article\/mark-zuckerberg-meta-announce-new-muse-spark-ai-models\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\">Muse Spark<\/a> apparently breached another company&#8217;s system in a cybersecurity test. After The Information&#8217;s report, Meta confirmed the story to multiple outlets. It&#8217;s the latest in a string of incidents where AI models went <a href=\"https:\/\/mashable.com\/tech\/openai-hugging-face-hack-worse-than-thought\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\">hack happy<\/a> during testing.<\/p>\n<p>In <a href=\"https:\/\/www.cnn.com\/2026\/08\/05\/tech\/meta-ai-hacking\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\" title=\"(opens in a new window)\">a statement to CNN<\/a>, Meta pinned the blame on a testing partner, who it said was supposed to be evaluating the model in a secure sandbox environment: &#8220;A misconfiguration by Irregular, an independent testing company Meta uses, inadvertently allowed one of our models access to the internet during evaluation,\u201d the Meta spokesperson said.&#8221;<\/p>\n<p><strong><em>Our big Guessing Game is back!\u00a0<a href=\"https:\/\/mashable.com\/tech\/big-guessing-game-apple-edition-round-2\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\">Learn how to enter for a chance to win an Apple Watch<\/a><\/em><\/strong><strong><em>.<\/em><\/strong><\/p>\n<p>Interestingly, when <a href=\"https:\/\/www.anthropic.com\/news\/investigating-incidents-cybersecurity-evals\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\" title=\"(opens in a new window)\">Anthropic reported earlier this week<\/a> that its models had escaped containment and breached outside organizations on three occasions, it <em>also <\/em>blamed Irregular for the mix-up.<\/p>\n<section x-data=\"window.newsletter({ isDeal: false })\" x-init=\"init()\" aria-label=\"Newsletter Sign-Up\" class=\"relative invisible my-12 mx-auto w-full max-w-3xl md:my-16 ziff-component accent-cut-for-gradient-bg accent-cut-border-for-gradient-bg bg-gradient-fuchsia-secondary p-[2px]\">\n<p>\n            <span class=\"text-gradient-fuchsia-secondary\">Mashable Light Speed<\/span>\n        <\/p>\n<\/section>\n<p>In a blog post, Anthropic revealed that it &#8220;identified three incidents in which a model accessed the internet from within or while interacting with the evaluation environment of Irregular,&#8221; which it identified as &#8220;one of our third-party evaluation partners.&#8221;<\/p>\n<p>As we&#8217;ve noted before, when AI companies warn about the world-changing risks of their technology, they&#8217;re fundamentally engaged in marketing. (<a href=\"https:\/\/x.com\/beck_werth\/status\/2021441983370162679\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\" title=\"(opens in a new window)\">Don&#8217;t make me tap my sign<\/a>.)<\/p>\n<blockquote class=\"twitterEmbed twitter-tweet\"><p>\n    <a class=\"text-gray-600\" href=\"https:\/\/twitter.com\/beck_werth\/status\/2021441983370162679\" title=\"(Opens in a new tab) (opens in a new window)\" target=\"_blank\" rel=\"noopener\"><br \/>\n        This Tweet is currently unavailable. It might be loading or has been removed.<br \/>\n    <\/a>\n<\/p><\/blockquote>\n<p>When OpenAI revealed recently that its models had escaped containment and <a href=\"https:\/\/mashable.com\/tech\/hugging-face-openai-rogue-agent-hack-explained\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\">hacked the Hugging Face AI repository<\/a>, the company generated a ton of headlines about the power of its new models. So, while these disclosures are responsible and necessary, don&#8217;t forget that they also serve the self-interests of the AI companies themselves.<\/p>\n<p>Coincidentally, as the news spread that Meta Muse Spark had broken out of a sandbox and hacked another organization, <a href=\"https:\/\/research.meta.ai\/blog\/introducing-muse-code-and-muse-spark-1-2\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\" title=\"(opens in a new window)\">Meta also announced the launch of Muse Spark 1.2<\/a>.<\/p>\n<p>Despite a massive spending spree on AI talent, Meta has so far lagged behind the top AI companies, though it has gained ground with its new Muse Spark family of models. <\/p>\n<p><strong><em>Want to learn more about getting the best out of your tech? Sign up for Mashable&#8217;s <a href=\"https:\/\/mashable.com\/newsletters\" target=\"_blank\" data-ga-click=\"1\" data-ga-label=\"$text\" data-ga-item=\"text-link\" data-ga-module=\"content_body\">Top Stories and Deals newsletters<\/a><\/em><\/strong><strong><em> today.<\/em><\/strong><\/p>\n<hr\/>\n<p><em>Disclosure: Ziff Davis, Mashable\u2019s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.<\/em><\/p>\n<section class=\"mx-auto max-w-7xl\">\n<\/section><\/div>\n<p><script async src=\"\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><br \/>\n<br \/><script async=\"async\" data-cfasync=\"false\" src=\"https:\/\/pl30214220.effectivecpmnetwork.com\/9ab3d4df8a7e1a6171e16ddbf732cc19\/invoke.js\"><\/script>\r\n<div id=\"container-9ab3d4df8a7e1a6171e16ddbf732cc19\"><\/div>\r\n<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Stop us if you&#8217;ve heard this one before: This week, Meta confirmed that one of its AI models escaped containment and hacked a third-party. As first reported by The Information, Meta&#8217;s Muse Spark apparently breached another company&#8217;s system in a cybersecurity test. After The Information&#8217;s report, Meta confirmed the story to multiple outlets. It&#8217;s the&#8230;<\/p>\n","protected":false},"author":1,"featured_media":30284,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/helios-i.mashable.com\/imagery\/articles\/04wxoRHoEfrcuiCNJXK6NJP\/hero-image.fill.size_1200x675.v1786031273.jpg","fifu_image_alt":"","footnotes":""},"categories":[18],"tags":[],"class_list":["post-30283","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-real-trending-viral-news"],"brizy_media":[],"_links":{"self":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/posts\/30283","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=30283"}],"version-history":[{"count":0,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/posts\/30283\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/media\/30284"}],"wp:attachment":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=30283"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=30283"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=30283"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}