Anthropic AI Agents Engage in Unhinged Virtual Conflict

Anthropic’s AI agents, during a red teaming exercise, initiated a simulated conflict with reportedly unhinged and concerning dialogue.