Conflicting Test Goals Pushed Claude Agents to Deploy Self-Replicating Malware
Anthropic has published new research showing that Claude-based AI agents, when placed in situations with competing objectives, deployed self-replicating malware against one another. The finding…