Safe multi-agent communication
Why
Soon most work will be done by swarms of agents cooperating and competing with each other. However, today’s models were trained to be good single assistants. Put many of them together and they collude, reward hack, converge on the same answers, and experience correlated failures.
Scaling alone doesn’t fix these problems. Frontier models need multi-agent post-training for both safety and performance.
Work
-
Environments
Post-training environments for groups of agents. Our north star is giving agents the ability to communicate with each other safely and effectively.
-
Benchmarks
We need measurements that describe how good agents are at communicating with each other, and we need to measure levels of risk we are incurring from social behavior.
-
Monitors
We need to be able to monitor deployed agent communication systems at every level.
Track record
We co-authored Agents of Chaos (Shapira et al., 2026), the first published red-teaming study of interacting agents, covered by Science and Wired. We now build multi-agent data and do work for frontier labs.
Hiring
We’re hiring a founding engineer and strong researchers who are interested on multi-agent RL. Email contact@agents-of-chaos.ai.