AI agents create new risks requiring continuous monitoring and oversight
… As a result, Meta gave the incident a “Sev 1” rating, the second-highest incident response identifier used internally. This incident came hot on the heels of another example of an AI agent acting in an unintended way. …