{"id":125106,"date":"2026-07-31T01:08:11","date_gmt":"2026-07-31T01:08:11","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/125106\/"},"modified":"2026-07-31T01:08:11","modified_gmt":"2026-07-31T01:08:11","slug":"anthropic-says-ai-models-hacked-three-firms-during-tests","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/125106\/","title":{"rendered":"Anthropic says AI models hacked three firms during tests"},"content":{"rendered":"<p class=\"mb-4 text-lg md:leading-8 break-words\">US tech company Anthropic says three of its artificial intelligence (AI) models broke out of isolated test environments and hacked into the systems of three other firms during a cybersecurity exercise.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">It comes just days after rival OpenAI said that its models had <a data-ylk=\"slk:breached the systems of other companies;elm:context_link;itc:0;sec:content-canvas;\" data-yga=\"{&quot;yLinkElement&quot;:&quot;link&quot;,&quot;yLinkElementType&quot;:&quot;article_link&quot;}\" href=\"https:\/\/www.bbc.com\/news\/articles\/c2el319vzr3o?xtor=AL-72-%5Bpartner%5D-%5Byahoo.north.america%5D-%5Blink%5D-%5Bnews%5D-%5Bbizdev%5D-%5Bisapi%5D\" rel=\"noopener noreferrer nofollow\" target=\"_blank\">breached the systems of other companies<\/a>, including Hugging Face &#8211; a hub for AI tools.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">The announcement prompted Anthropic to check whether its own models had carried out similar attacks. It says it uncovered three cases that have since been reported to the affected companies.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">Anthropic, which did not name the firms, urged other AI labs to perform similar reviews to better understand the risks of their models&#8217; capabilities.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">Anthropic said <a data-ylk=\"slk:in a statement;elm:context_link;itc:0;sec:content-canvas;\" data-yga=\"{&quot;yLinkElement&quot;:&quot;link&quot;,&quot;yLinkElementType&quot;:&quot;article_link&quot;}\" href=\"https:\/\/www.anthropic.com\/news\/investigating-incidents-cybersecurity-evals\" rel=\"noopener noreferrer nofollow\" target=\"_blank\">in a statement<\/a> that it reviewed more than 140,000 tests to find evidence that Claude &#8211; its family of AI models &#8211; could access the internet from testing environments that were designed to be sealed off.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">The tests include so-called &#8220;capture-the-flag&#8221; evaluations in which Claude was tasked with obtaining information by breaching other systems &#8211; a common way that experts assess a model&#8217;s hacking capabilities.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">A &#8220;miscommunication&#8221; on systems run by Anthropic and its testing partner left the models with live internet access, allowing them to breach other systems, the San Francisco-based firm said.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">Anthropic added that it is &#8220;approaching the fixes as if the responsibility were ours alone.&#8221;<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">A string of AI-driven cyberattacks has fuelled calls for tighter safeguards and oversight of the technology, over concerns about the risks posed by increasingly powerful autonomous systems.<\/p>\n<p class=\"mb-4 text-lg md:leading-8 break-words\">US President Donald Trump said on Wednesday that Washington is <a data-ylk=\"slk:considering measures to rein in artificial intelligence tools;elm:context_link;itc:0;sec:content-canvas;\" data-yga=\"{&quot;yLinkElement&quot;:&quot;link&quot;,&quot;yLinkElementType&quot;:&quot;article_link&quot;}\" href=\"https:\/\/www.bbc.com\/news\/articles\/c20dppq3y90o?xtor=AL-72-%5Bpartner%5D-%5Byahoo.north.america%5D-%5Blink%5D-%5Bnews%5D-%5Bbizdev%5D-%5Bisapi%5D\" rel=\"noopener noreferrer nofollow\" target=\"_blank\">considering measures to rein in artificial intelligence tools<\/a> after recent cybersecurity incidents.<\/p>\n","protected":false},"excerpt":{"rendered":"US tech company Anthropic says three of its artificial intelligence (AI) models broke out of isolated test environments&hellip;\n","protected":false},"author":2,"featured_media":125107,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[53,25,1160,37987],"class_list":["post-125106","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-anthropic","tag-artificial-intelligence","tag-models","tag-test-environments"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/125106","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=125106"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/125106\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/125107"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=125106"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=125106"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=125106"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}