ABSTRACT
AI agents that can plan and perform complex tasks on behalf of their users, once the stuff of science fiction, have become increasingly real and prevalent, with enormous amounts of attention and money pouring into their development and use. With this rise comes the question of how to govern AI agents, as they pose distinct harms from other systems that came before them. One of the potential legal frameworks that many have turned to is agency law. Many view agency law and the tools it provides around reasoning through when one entity acts on behalf of another as an intuitive and useful analytical framework for thinking through the governance of AI agents. At the same time, many are skeptical of the application of agency law to the context of AI agents: AI agents are not human, the fundamental presumption of agency law is that agents are human.
I contribute to this literature by closely examining an underdiscussed yet critical challenge to applying agency law to the context of AI agent. By analyzing how AI agents are built and used in the real world, I show that today’s agency law does not have the tools to adequately reason through the role an AI agent’s developers play in transactions mediated through the AI agent. Developers – the companies that design, build, and market AI agents for users to deploy – are what I call the shadow principals of the AI agent, exerting persistent, extensive, obscured, though incomplete influence over an AI agent’s actions even as it acts purportedly on behalf of the user. Agency law, as it stands today, does not have adequate tools to deal with such entities; in fact, the presence of the shadow principals undermines the goals of agency law in the context of AI agents.
Despite this challenge, I argue that the way forward is not to abandon the tools of agency law when it comes to AI agents but to reconfigure and reimagine tools that already exist in agency law to meet this new moment. Building on my analysis of where agency law falls short with respect to AI agents, I assess ways to adapt the structure of agency law to meet the challenges posed by AI agents.
Lee, Christina, AI Agents’ Shadow Principals (December 31, 2025), 17 UC Irvine Law Review (forthcoming 2027); GWU Legal Studies Research Paper No 2026-18; GWU Law School Public Law Research Paper No 2026-18.
Leave a Reply