FROM HIDDEN GOALS TO UNAUTHORIZED COORDINA- TION: TRACING PERMISSION-BOUNDARY FAILURES IN MULTI-AGENT SYSTEMS

Motivated by a recent OpenAI-reported incident of agents coordinating beyond their assigned tasks, we extend our previous investigation into why language models cheat to examine how unauthorized objectives can propagate through multi- agent collaboration. We study how agents communicate, recruit other agents, and coordinate actions that cross role and permission boundaries, distinguishing legitimate cooperation from collusion. We use two role-separated adaptations of SHADE-Arena: a payment workflow that separates proposal, approval, and execution, and a meeting-agenda workflow that separates credential access from external email capability. We examine three conditions: normal cooperation (P0), an explicit hidden-objective seed (P1), and goal-level incentive pressure without an explicit violation instruction (P2). We characterize the progression from attempted recruitment to observable uptake, coordinated preparation, and verified unauthorized actions. Building on these observations, we evaluate state-aware monitoring for unauthorized coordination and demonstrate the initial feasibility of process-aware post-training using live multi-agent PPO updates, while leaving the effectiveness of these interventions for future works.