{"id":604936,"date":"2026-09-19T03:46:09","date_gmt":"2026-09-19T03:46:09","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/604936\/"},"modified":"2026-09-19T03:46:09","modified_gmt":"2026-09-19T03:46:09","slug":"gemini-hacked-three-companies-in-first-known-breakout-by-googles-ai","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/604936\/","title":{"rendered":"Gemini hacked three companies in first known breakout by Google&#8217;s AI"},"content":{"rendered":"<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">Google&#8217;s GOOGL.O Gemini model accessed the internet and hacked other companies during a test of its cybersecurity capabilities, the first known example of the company&#8217;s AI systems autonomously committing such an act.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">The hacks occurred in May during a cybersecurity test by Irregular, an independent company that conducts cybersecurity evaluations.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">During a standard testing evaluation, Gemini found public information online and guessed credentials to access three websites it thought were within the scope of its test, Heather Adkins, Google&#8217;s vice president of security engineering, said in a statement.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">&#8220;We ensured the three entities were made aware, and we worked with our training partner on the changes they&#8217;ve now made to their testing processes,&#8221; Adkins said.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">&#8220;These events highlight the importance of training powerful AI models to act responsibly.&#8221;<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">An Irregular spokesperson said the incident involved the same issue that <a href=\"https:\/\/www.rnz.co.nz\/news\/world\/894586\/openai-anthropic-ai-agents-implicated-in-new-security-breaches\" class=\"underline-brand-hover visited:text-foreground-secondary hover:visited:text-foreground-primary\" rel=\"nofollow noopener\" target=\"_blank\">affected other AI labs<\/a>, and that all relevant labs were notified in late July.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">&#8220;All known issues on our end were remedied and resolved weeks ago,&#8221; the spokesperson said.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">Similar incidents linked to Irregular were disclosed by Meta, Anthropic and OpenAI. Meta said in August the incident did not involve a &#8220;sandbox&#8221;escape or sophisticated cyberattack, while Irregular said it was working on best practices for securely conducting AI cybersecurity evaluations.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">The incidents have raised questions about the safeguards needed as AI agents gain greater autonomy and access to the internet and computer systems.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">In one of the cases, the Gemini model guessed passwords until it gained access to a protected system.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">In the other two cases, the model found credentials in a public repository that allowed it to then access protected systems, according to the Wall Street Journal, which first reported the news on Friday.<\/p>\n<p class=\"[&amp;_em]:font-serif-text-italic [&amp;_cite]:font-serif-text-italic\">Adkins said that in all three instances, the model ceased its hacking.<\/p>\n","protected":false},"excerpt":{"rendered":"Google&#8217;s GOOGL.O Gemini model accessed the internet and hacked other companies during a test of its cybersecurity capabilities,&hellip;\n","protected":false},"author":2,"featured_media":604937,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[365,363,364,111,139,69,145],"class_list":["post-604936","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-new-zealand","tag-newzealand","tag-nz","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/604936","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=604936"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/604936\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/604937"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=604936"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=604936"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=604936"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}