On AI:AM

Lewis Hammond

Research Director, Cooperative AI Foundation

Lewis Hammond is Research Director at the Cooperative AI Foundation, where he works on how AI systems interact with each other and with people. He is first author of "Multi-Agent Risks from Advanced AI" (February 2025), the roughly 45-author report that set out three failure modes for interacting AI systems — miscoordination, conflict and collusion — alongside seven risk factors including emergent agency and multi-agent security. His later work includes "Causal Foundations of Collective Agency," which gives a behavioural criterion for when a group of agents is best modelled as a single agent, and "Neural Interactive Proofs," on protocols that let a bounded verifier extract correct answers from a more powerful untrusted prover. He holds a DPhil in computer science from the University of Oxford and is an affiliate of the Centre for the Governance of AI.