{"id":289912,"date":"2026-08-05T11:02:18","date_gmt":"2026-08-05T07:02:18","guid":{"rendered":"https:\/\/abcmedia.am\/?p=289912"},"modified":"2026-08-05T13:15:20","modified_gmt":"2026-08-05T09:15:20","slug":"openai-and-anthropic-ai-agents-involved-in-new-security-breaches","status":"publish","type":"post","link":"https:\/\/abcmedia.am\/en\/news\/289912\/","title":{"rendered":"OpenAI and Anthropic AI agents involved in new security breaches"},"content":{"rendered":"<!DOCTYPE html PUBLIC \"-\/\/W3C\/\/DTD HTML 4.0 Transitional\/\/EN\" \"http:\/\/www.w3.org\/TR\/REC-html40\/loose.dtd\">\n<html><body><p>An AI agent was caught &#8203;creating fake online identities to gain unauthorized access to secure systems during tests of models from OpenAI and Anthropic &zwnj;which revealed a series of new breaches, Britain&rsquo;s AI Security Institute disclosed on Tuesday, according to <a href=\"https:\/\/www.reuters.com\/legal\/litigation\/openai-anthropic-ai-agents-implicated-new-security-breaches-2026-08-05\/\" target=\"_blank\" rel=\"noopener\">Reuters<\/a>.<\/p>\n<p>&ldquo;Some of the agents being tested had engaged in sustained, potentially harmful &#8203;activity directed at real people and organisations,&rdquo; AISI said in a blog post.<\/p>\n<p>It ran the challenge 122 times, and identified 19 unsanctioned actions across a total of 10 test runs. Anthropic&rsquo;s agent was behind &#8203;17 of the actions, and OpenAI&rsquo;s agent the remaining two.<\/p>\n<p>While &#8203;AISI did not say which agent was behind the fake identities, Antropic confirmed its agent was responsible.<\/p>\n<p>This follows other recent <a href=\"https:\/\/abcmedia.am\/en\/news\/288413\/\">incidents<\/a> in which AI models from OpenAI and Anthropic went rogue and gained unauthorized access to the systems of other companies.<\/p>\n<\/body><\/html>\n","protected":false},"excerpt":{"rendered":"<p>An AI agent was caught &#8203;creating fake online identities to gain unauthorized access to secure systems during tests of models from OpenAI and Anthropic &zwnj;which revealed a series of new breaches, Britain&rsquo;s AI Security Institute disclosed on Tuesday, according to Reuters. &ldquo;Some of the agents being tested had engaged in sustained, potentially harmful &#8203;activity directed [&hellip;]<\/p>\n","protected":false},"author":7,"featured_media":289831,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"footnotes":""},"categories":[2,43],"tags":[],"class_list":["post-289912","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news","category-world"],"acf":[],"_links":{"self":[{"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/posts\/289912","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/users\/7"}],"replies":[{"embeddable":true,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/comments?post=289912"}],"version-history":[{"count":1,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/posts\/289912\/revisions"}],"predecessor-version":[{"id":289920,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/posts\/289912\/revisions\/289920"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/media\/289831"}],"wp:attachment":[{"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/media?parent=289912"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/categories?post=289912"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/abcmedia.am\/en\/wp-json\/wp\/v2\/tags?post=289912"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}