A free 3-day research hackathon on failures that emerge when many AI agents interact: collusion, coordinated deception, attacks that spread from agent to agent, and agents that rename themselves to dodge oversight. Teams build detectors for coordinated behaviour in real multi-agent logs, reputation systems that colluding agents can’t game, agent IDs that survive copying, and safe sandboxes for testing. Experience in trust and safety, fraud and integrity is especially welcome. Online worldwide, with an in-person hub at AE Studio in Los Angeles. $4,000 in prizes, plus $4,000 for the LA hub.