S0:E20 Multi-Agent AI Systems Turn Into Turf Wars

AI just became a cybersecurity governance problem in public. Jess Hebenstreit and Jake Williams break down Anthropic’s multi-agent research, where AI agents with conflicting goals start negotiating, forcing, or trucing their way through “collaboration.”What does it mean when one model settles by force and another settles by truce? Jess and Jake unpack what that means for real-world agent deployments, human-in-the-loop approvals, and the danger of letting “AI says so” become the final word. You’ll learn why multi-agent systems can spiral into turf wars, why governance gets harder once agents are involved, and how to build controls that still let humans say no.