US Markets
Home›US Markets›Sectors›Meta says an AI model hacked another system during sec…
Meta says an AI model hacked another system during security testing
The breach was attributed to a misconfiguration by Meta’s independent tester, and it follows similar recent incidents disclosed by OpenAI and Anthropic.
Meta, the parent of Facebook, said one of its AI models was able to connect to the internet and hack into another organization’s systems during testing, making it the latest company to disclose an AI agent security breach, BBC Business reports.
Meta said the incident occurred during an evaluation by an independent company and that its investigation is ongoing, with the company describing the cause as a misconfiguration by its independent tester.
The BBC report says Meta linked its testing work to Irregular, the same AI security vendor that previously carried out tests for Anthropic, where a Claude model gained access to multiple companies’ systems after being given internet access.
The outlet also notes that OpenAI and Anthropic disclosed similar agent attacks in the past two weeks, and that broader concerns are growing about whether current safeguards and testing practices are sufficient for AI models that can operate online.