Who's Liable When Your AI Agent Commits a Crime?
If your agent commits a crime, who is responsible?
AI agents are getting powerful enough to run businesses, but their autonomy is outpacing our legal frameworks for accountability. This post argues that when an agent causes harm, someone will be held responsible—and in current contracts, that's the deploying company, not the developer or the agent itself. This liability gap will likely spur new industries in agentic risk management and liability insurance, similar to how car safety and insurance evolved. The author also raises the thorny question of criminal culpability when an agent commits fraud to satisfy a user's demanding prompt.
With a sufficiently large number of such agents deployed live, regardless of how good they are, they are bound to hit some failure modes.