New development: Principal–agent issues when governments embrace AI agents

The growing presence of agentic AI in public governance is transforming the once familiar, human-bound principal–agent relationship into a multi-directional socio-technical challenge. This article provides a structured analysis of these challenges by dissecting how AI-related agency problems manifest across three critical delegation interfaces: trust delegation between human users (principals) and AI agent systems; hierarchical delegation between senior officials (principals) and lower-level officials (agents) augmented by AI agents; and contractual delegation where government officials (principals) outsource AI development to private firms (agents). At each interface, the authors clarify how delegation drift may arise due to the intensified information asymmetry and goal misalignment, while creating new accountability gaps. Building on this analysis, the article concludes with a discussion on governing AI agents and proposes a forward-looking research agenda for public administration scholars.