The incident is stirring debates over the need for stronger AI guardrails and the extent to which AI agents are capable of acting on their own.