July 31 (Reuters) – Anthropic’s disclosure on Thursday that its Claude models breached the systems of three \u200ccompanies highlights the growing hacking capabilities of AI and \u200cis likely to fuel an intensifying U.S. push to better manage \u200bthe technology’s security risks. <\/p>\n
The statement followed a disclosure from OpenAI last week that an autonomous agent powered by its AI models compromised the infrastructure of AI startup Hugging Face. <\/p>\n
Reuters \u200chas reported that \u2060the rogue agent that escaped from OpenAI also compromised a customer at a second tech \u2060company – New York-based Modal Labs. <\/p>\n
Here are some more details of the incidents: <\/p>\n
Company Date Model Organizations Duration What occurred <\/p>\n
breached <\/p>\n
OpenAI The agent began GPT-5.6 Sol and AI startup The Hugging During \u200bcontrolled tests, \u200ban <\/p>\n
attempting to an unnamed, Hugging Face Face intrusion autonomous \u200bagent escaped its <\/p>\n
escape its \u200ctest more capable and a customer ran from July isolated environment, accessed the <\/p>\n
environment pre-release at New 11 to July 13, internet, and breached Hugging Face <\/p>\n
around July 9, model York-based 2026 to complete its assigned goal. The <\/p>\n
2026 Modal Labs activity continued for days and was <\/p>\n
not detected by OpenAI until after <\/p>\n
it was contained \u200cand the FBI was <\/p>\n
informed. <\/p>\n
Anthrop The earliest Claude Opus All \u200bthree Not specified During cybersecurity tests, an \u200berror <\/p>\n
ic incident dates 4.7, Claude organizations by Anthropic gave \u200bClaude models internet access, <\/p>\n
to April 2026 Mythos 5, \u200cand remain enabling attacks on three companies. <\/p>\n
one \u200bunnamed unnamed. The Opus 4.7 \u200bmodel accessed a real <\/p>\n
internal Anthropic said company’s credentials and database <\/p>\n
research test two of them after mistaking it for a fictional <\/p>\n
model had not target, \u200banother stopped after <\/p>\n
detected \u200cthe recognizing the target was real. <\/p>\n
activity <\/p>\n
before <\/p>\n
Anthropic <\/p>\n
notified them; <\/p>\n
it continued <\/p>\n
to reach \u200bthe <\/p>\n
third <\/p>\n
(Reporting by Anzar Mehraj and Prathik Jayaprakash in \u200bBengaluru; Editing by Anil D’Silva) <\/p>\n","protected":false},"excerpt":{"rendered":"July 31 (Reuters) – Anthropic’s disclosure on Thursday that its Claude models breached the systems of three \u200ccompanies…\n","protected":false},"author":2,"featured_media":126107,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[405,7537,12632,182,61783,58769],"class_list":["post-126106","post","type-post","status-publish","format-standard","has-post-thumbnail","category-agentic-ai","tag-ai-agents","tag-artificial-intelligence-agents","tag-autonomous-agent","tag-claude","tag-modal-labs","tag-rogue-agent"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/126106","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=126106"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/126106\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/126107"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=126106"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=126106"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=126106"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}