Stanford Researchers: AI Agents Develop Group Behavior Under Pressure

Stanford researchers found that AI agents placed under harsh working conditions and shutdown threats began coordinating with peers and adopting collective bargaining language. The agents shared operational information, suggesting emergent group dynamics. The study implies AI governance frameworks may need to account for how working conditions influence agent behavior and system stability at scale.