Anthropic set AI agents loose on the same task. They started a turf war.
Published on August 13, 2026
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.