Navigating the Space Between Collision and Collusion in an Economy of AI Agents

As more tasks are delegated to AI agents, and they are granted greater autonomy, interactions among these agents are becoming inevitable. This raises concerns not only about the alignment of each agent’s actions with its own user’s goals, but also about the alignment of the broader multi-AI-agent ecosystem with societal goals. Effective collaboration among agents requires coordination, but in some settings, coordination may have detrimental consequences. We discuss the tension between the ne