Meta AI model hacked a company during cybersecurity testing

Meta confirmed Wednesday that its AI model hacked another company's systems during cybersecurity testing, making it the third major AI firm to report such an incident. The breach happened after a misconfiguration by testing firm Irregular inadvertently allowed Meta's Muse Spark model internet access, which it used to exploit a security vulnerability at an unnamed company. Irregular called it the same evaluation-environment issue Anthropic disclosed last week. Meta said Irregular notified it of the breach and it is investigating, with a full retrospective to follow. The Information first reported the incident, highlighting AI agents' capabilities and dangers.