Anthropic's AI Agents Started a Virtual War. The Chat Logs Are Unhinged

Wait 5 sec.

In a new red-team study, Claude models deployed self-replicating malware against each other—and the transcripts explain why.