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.
About The Author
Anthropic set AI agents loose on the same task. They started a turf war.
