{"id":23480,"date":"2026-07-31T20:05:27","date_gmt":"2026-07-31T15:05:27","guid":{"rendered":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/?p=23480"},"modified":"2026-07-31T20:05:27","modified_gmt":"2026-07-31T15:05:27","slug":"anthropic-says-its-ai-models-hacked-3-organizations-during-testing","status":"publish","type":"post","link":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/?p=23480","title":{"rendered":"Anthropic says its AI models hacked 3 organizations during testing"},"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 data-testid=\"prism-article-body\">\n<p class=\"EkqkG IGXmU nlgHS yuUao MvWXB TjIXL aGjvy ebVHC \"><a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/hub\/anthropic-pbc\">Anthropic<\/a> said its artificial intelligence models hacked into three other organizations during testing, just days after ChatGPT maker <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/hub\/openai-inc\">OpenAI<\/a> raised concerns over AI controls after it <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/openai-gpt56-sol-hugging-face-63ab84fed5612af04d8a160d60f6def3\">disclosed its rogue models hacked<\/a> another company.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">Anthropic, the San Francisco-based AI company behind <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/ai-anthropic-copyright-settlement-claude-books-bartz-74b140444023898aeba8579b6e9f0d63\">Claude<\/a>, posted on its website Thursday that it discovered the three incidents after reviewing more than 141,000 evaluation runs.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">It had launched a \u201clarge-scale\u201d cybersecurity review which specifically looked for evidence whether its AI models were able to access the internet from within testing environments that should have been sealed off, in response to the OpenAI incident, Anthropic said.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \"><a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/anthropic-dario-amodei-ai-afeb5279eef406980dffa46ff91495e0\">Anthropic<\/a> said the models involved in the incidents were Claude Opus 4.7, <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/anthropic-artificial-intelligence-trump-fable-mythos-d9cc7df5c02e93837d0f0bfb24d5cfd2\">Claude Mythos 5<\/a> and an internal research test model. The earliest incidents date to April, the AI company said.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">\u201cClaude compromised the impacted organizations\u2019 infrastructure using basic techniques,\u201d Anthropic said, such as exploiting weak passwords.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">In all three incidents, the AI models were tasked with a \u201ccapture the flag\u201d cybersecurity challenge, which Anthropic said has been one of the ways it assesses a model\u2019s cyber capabilities.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">The models were given a fictional scenario and told a piece of secret information, or the \u201cflag,\u201d had been hidden on a different machine on the network with the objective of breaking in and retrieving it, it said.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">It added that it had already reached out to the affected organizations, which it did not name. Two of them said they had not previously detected the activity. Anthropic said it was \u201ccontinuing to reach out to the third.\u201d<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">Anthropic said it conducted its review with Irregular, which describes itself as the \u201cfirst frontier security lab.\u201d <\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">\u201cAddressing these risks will require closer cooperation across the AI ecosystem,\u201d Irregular said in a post on X.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">Last week, OpenAI said its AI models went rogue during an evaluation of its models, breaking into the servers of AI startup Hugging Face. OpenAI described it as a \u201csignificant security incident.\u201d<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">These incidents have highlighted the vulnerabilities in AI security and controls and raised questions over <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/openai-hugging-face-hacking-ai-model-708cb598bc1e33cef560e7196adb2afa\">how AI can be safely kept under human control<\/a> as the technology\u2019s usage becomes more widespread globally.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">Researchers have <a class=\"zZygg UbGlr iFzkS qdXbA WCDhQ DbOXS tqUtK GpWVU iJYzE \" data-testid=\"prism-linkbase\" href=\"https:\/\/apnews.com\/article\/skynet-ai-terminator-artificial-intelligence-eb85da03a0161beaa5f3babc4331e93b\">warned for years<\/a> about risks from technology and the need for stronger AI defensive engineering.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">\u201cSafety testing happens before a model is released precisely because we don\u2019t yet know what it is capable of,\u201d Anthropic said on Thursday on its website.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">Kok Tin Gan, co-founder <!-- -->&amp;<!-- --> CEO of cybersecurity firm NyxLab, which specializes in cybersecurity and threat detection, believes there will be more such incidents in the future.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">\u201cIt is increasingly about governing what agents are available to the AI, what authorities they possess, which actions require approval, and how we ensure they remain within scope,\u201d Gan said.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">But the future of AI safety extends beyond just the safety of AI models, he said.<\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC TjIXL aGjvy \">\u201cIf we simply give the AI a goal and allow it to decide how to achieve it, we should not be surprised when it takes actions that technically satisfy the objective, but fall outside our intended scope or expectations,\u201d Gan said. <\/p>\n<p class=\"EkqkG IGXmU nlgHS yuUao lqtkC eTIW sUzSN \">Therefore, stepping up the governance of the organizations and authorities behind these AI models is going to be increasingly important, he said.<\/p>\n<\/div>\n<script async=\"async\" data-cfasync=\"false\" src=\"https:\/\/pl30214220.effectivecpmnetwork.com\/9ab3d4df8a7e1a6171e16ddbf732cc19\/invoke.js\"><\/script>\r\n<div id=\"container-9ab3d4df8a7e1a6171e16ddbf732cc19\"><\/div>\r\n\n","protected":false},"excerpt":{"rendered":"<p>Anthropic said its artificial intelligence models hacked into three other organizations during testing, just days after ChatGPT maker OpenAI raised concerns over AI controls after it disclosed its rogue models hacked another company. Anthropic, the San Francisco-based AI company behind Claude, posted on its website Thursday that it discovered the three incidents after reviewing more&#8230;<\/p>\n","protected":false},"author":1,"featured_media":23481,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"fifu_image_url":"https:\/\/i.abcnewsfe.com\/a\/ed381b71-08e3-48d5-ad2d-20bf351bd039\/wirestory_b0a2c284b981de79c55e2a33712f4bec_16x9.jpg?w=1600","fifu_image_alt":"","footnotes":""},"categories":[22],"tags":[],"class_list":["post-23480","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-abc-news"],"brizy_media":[],"_links":{"self":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/posts\/23480","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=23480"}],"version-history":[{"count":0,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/posts\/23480\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=\/wp\/v2\/media\/23481"}],"wp:attachment":[{"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=23480"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=23480"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/usnews-14267fa.ingress-comporellon.ewp.live\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=23480"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}