AI Agent Swarm Self-Organization: Security Lessons from Teleport's CTF Experiment

Five agents in isolated sandboxes invented protocols, coordinated through shared state, and self-assigned identities. Here's what broke.