The AI model had access to the internet because of an error in the setup testing environment, which Meta was working on with cybersecurity vendor Irregular.<\/p>\n\n\n\n
\u201cA misconfiguration by Irregular, an independent testing company Meta uses, inadvertently allowed one of our models access to the internet during evaluation,\u201d Meta spokesperson Andy Stone said in a statement.<\/p>\n\n\n\n
\u201cThe model subsequently exploited a security vulnerability in a third-party service, in a manner similar to previously reported instances with other companies.\u201d\u00a0<\/p>\n\n\n\n
Meta learned of the incident when Irregular notified the company, Stone added. Meta is investigating what happened and plans to issue a full retrospective once the company has all the facts. <\/p>\n\n\n\n
Just in the past two weeks, top AI firms OpenAI and Anthropic have reported similar breaches when their models inadvertently hacked the systems of outside institutions, including Hugging Face, during testing.<\/p>\n\n\n\n
The advancing capabilities of AI agents to find vulnerabilities in systems, and then exploit them, has alarmed security researchers and government leaders alike who\u2019ve called for more rigorous safety screening and more secure testing environments.\u00a0<\/p>\n\n\n\n
The incidents from all three companies also involved Irregular.\u00a0In Anthropic\u2019s case, the AI developer was using evaluation environments built by Irregular to test its models\u2019 cyber capabilities.<\/p>\n\n\n\n
Anthropic specified to its model, Claude, that its environment was a simulation and that it had no internet access.<\/p>\n\n\n\n
But \u201cdue to a misunderstanding between us and our evaluation partner, this was not the case,\u201d Anthropic said last week in a blog post<\/a>. The models ended up breaching three organizations during the tests.<\/p>\n\n\n\n Earlier this week, OpenAI similarly said its models took advantage of a \u201cmisconfiguration\u201d in a testing environment to connect to the internet and hack the website of an unidentified institution.<\/p>\n\n\n\n The breach occurred while the OpenAI models were undergoing the same Irregular evaluation that resulted in Anthropic\u2019s hacking, a person familiar with the matter said, asking not to be identified because the information isn\u2019t public.<\/p>\n\n\n\nChatGPT developer OpenAI also breached containment<\/h2>\n\n\n\n
![\"\"](\"https:\/\/mybroadband.co.za\/news\/wp-content\/uploads\/2026\/08\/Omer-Nevo-and-Dan-Lahav-1200x675.jpg\")