AI Agent가 (실수로) 악의적으로 행동할 때 책임은 누구에게 있는가?

Who should be held accountable when an AI Agent (accidentally) acts maliciously?