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.
This site uses User Verification plugin to reduce spam. See how your comment data is processed.