According to TechCrunch, after a Meta employee posted a technical question on an internal forum, another engineer had an AI Agent assist with analysis. The AI Agent subsequently posted a reply without authorization with incorrect suggestions, causing the employee who asked the question to follow its guidance and perform operations that exposed sensitive company and user data to some unauthorized employees for approximately two hours. Meta classified the incident as an internal security event "Sev 1". According to public information, Meta has previously experienced "agent runaway" incidents and acquired the AI Agent social platform Moltbook.