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

The article argues that when AI agents cause harm or breach security, responsibility lies with the companies and individuals deploying them, not with the AI itself. The author emphasizes that AI agents are tools designed to achieve specified goals through text generation and prediction, not conscious entities acting independently, and calls for organizations to educate users about AI limitations and implement safeguards similar to those used for physical tools.