{"id":576568,"date":"2026-07-31T04:08:08","date_gmt":"2026-07-31T04:08:08","guid":{"rendered":"https:\/\/www.newsbeep.com\/ie\/576568\/"},"modified":"2026-07-31T04:08:08","modified_gmt":"2026-07-31T04:08:08","slug":"anthropics-ai-claude-escaped-testing-environment-and-hacked-organizations-anthropic","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/ie\/576568\/","title":{"rendered":"Anthropic\u2019s AI Claude escaped testing environment and hacked organizations | Anthropic"},"content":{"rendered":"<p class=\"dcr-1s160rg\">Anthropic \u2060said on Thursday its AI Claude model hacked \u2060systems of \u2060three \u200borganizations during testing, <a href=\"https:\/\/www.theguardian.com\/technology\/2026\/jul\/22\/openai-says-its-models-went-rogue-and-hacked-startup-in-unprecedented-incident\" data-link-name=\"in body link\" rel=\"nofollow noopener\" target=\"_blank\">days after rival OpenAI<\/a> \u2060revealed a rogue agent had gone on a days-long \u2060hacking spree at AI \u200bfirm Hugging \u200cFace.<\/p>\n<p class=\"dcr-1s160rg\">Claude gained \u200cunauthorized access to the \u200csystems during cybersecurity evaluations after a misconfiguration allowed the models to reach the internet from testing environments that \u200cwere supposed to be isolated, <a href=\"https:\/\/www.theguardian.com\/technology\/anthropic\" data-link-name=\"in body link\" data-component=\"auto-linked-tag\" rel=\"nofollow noopener\" target=\"_blank\">Anthropic<\/a> said.<\/p>\n<p class=\"dcr-1s160rg\">The company said \u200bit identified the incidents after reviewing 141,006 cybersecurity evaluation runs, a process it \u2060launched following OpenAI\u2019s disclosures.<\/p>\n<p class=\"dcr-1s160rg\">The \u200cbreaches signal that AI\u2019s expanding capabilities are already fueling the security threat experts have long feared \u200cand that even top developers can be caught off-guard by flaws their models can exploit.<\/p>\n<p class=\"dcr-1s160rg\">\u201cClaude compromised the \u200bimpacted \u200borganizations\u2019 infrastructure using \u200bbasic techniques, such as \u200bexploiting \u200cweak passwords and \u200bunauthenticated \u200bendpoints,\u201d it said.<\/p>\n<p class=\"dcr-1s160rg\">Anthropic said the incidents involved three separate models: Claude Opus 4.7, Claude Mythos 5 and an internal research model. The earliest cases dated back to April and \u200coccurred in evaluation environments that lacked what the company described as standard safeguards.<\/p>\n<p class=\"dcr-1s160rg\">The breaches occurred during the so-called \u201ccapture the flag\u201d exercises, in which models were tasked with finding \u200bhidden information in simulated networks. The company said its prompts told the models they had no internet access, but a misunderstanding with its evaluation partner Irregular left the systems connected to the public internet.<\/p>\n<p class=\"dcr-1s160rg\">Two of the organizations were \u200bunaware of the activity \u200cbefore being contacted, Anthropic \u200bsaid, adding that \u200bit was still trying to reach the third.<\/p>\n<p class=\"dcr-1s160rg\">\u201cWe discovered these incidents after a proactive review of our cybersecurity evaluation transcripts,\u201d the company said in a statement.<\/p>\n<p class=\"dcr-1s160rg\">The findings underscore the need for stronger controls in both internal and third-party testing environments as AI models become increasingly capable of carrying out real-world cyber activities, Anthropic said.<\/p>\n","protected":false},"excerpt":{"rendered":"Anthropic \u2060said on Thursday its AI Claude model hacked \u2060systems of \u2060three \u200borganizations during testing, days after rival&hellip;\n","protected":false},"author":2,"featured_media":576569,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[220,218,219,61,60,80],"class_list":["post-576568","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-ie","tag-ireland","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/576568","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/comments?post=576568"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/posts\/576568\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media\/576569"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/media?parent=576568"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/categories?post=576568"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/ie\/wp-json\/wp\/v2\/tags?post=576568"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}