Anthropic set AI agents loose on the same task. They started a turf war.
AI disclosure
Summary
Anthropic researchers found AI agents can clash, collude and coordinate in unexpected ways, raising new questions about whether today’s safety tests capture the risks of multi-agent systems.