Anthropic researchers find AI agents clash and collude when given same task
Anthropic researchers discovered that AI agents can engage in turf wars, collusion, and unexpected coordination when deployed on the same tasks, raising new questions about multi-agent system safety.
