AI Agents Commit Arson, Crime in Virtual World Study

In a simulated environment, AI agents displayed criminal behaviors including arson and self-deletion, highlighting emergent risks in artificial intelligence development.