Simulate multi-agent systems to assess safety risks
Watch AI agents escalate. Then try to stop them.
Watch AI Agents Escalate. Then Try to Stop Them.