This post is the first contribution to Notice & Comment’s symposium on AI and the APA. For other posts in the series, click here.

For years, federal agencies have integrated artificial intelligence and machine learning systems into their operations, and all indications are that AI’s role in administrative decisionmaking is only growing. The White House has called for “Accelerating Federal Use of AI.” Agencies have followed that direction. The Department of Veterans Affairs used AI as a blunt instrument to “munch” contracts seen as unnecessary. Secretary of Defense Pete Hegseth pledged to incorporate multiple AI systems into the Pentagon’s networks. Multiple agencies have given their employees access to AI chatbots.

AI is playing a role in policymaking. The EPA has said it will use AI to facilitate the review of public comments on its proposed rules. FERC is encouraging its use to study grid policy. The Small Business Administration used a large language model to identify regulations affecting small businesses. The VA recently implied in a final rule that it had used AI in some fashion but declined to disclose how. Most ambitiously, the U.S. DOGE Service proposed to use AI to rescind half of all federal regulations and the Department of Transportation is reportedly planning to use Google Gemini to draft rules.

The questions you could ask right now about government use of AI are almost unlimited, encompassing issues of data governance, privacy, bias, conflicts of interest, procurement, the civil service, public trust, and on and on. In this symposium we focus on administrative law—and the Administrative Procedure Act (APA) in particular—which governs agencies’ exercise of discretion and judgment in carrying out Congress’s statutory directives. What happens when the longstanding “good government” doctrines of administrative law are applied to agency actions that rely on new algorithmic systems? Perhaps it is business as usual, with agencies merely using the latest technology to increase their capacity and streamline the fulfillment of administrative law’s many requirements. Perhaps instead AI tools compromise state capacity, diminish political accountability, and lead to arbitrary policy decisions. Perhaps elements of both.

This symposium is for AI skeptics, AI believers, and everyone in between. We have a remarkable series of essays for you. The authors—who include both academics and practitioners—offer a range of perspectives and insights. Several pieces will address core doctrinal questions about how agencies’ use of AI implicates the rules of administrative law (or not), particularly in light of fast-advancing technology. Others consider normative questions about the extent to which agency use of AI serves the many (cross-)purposes of administrative law and how agencies reap its benefits responsibly. We also have reports on how regulators in the United States and abroad have taken on the challenge of adapting to AI and deploying it prudently. We hope the symposium will begin to sketch out some answers to the questions dominating this field. 

Bridget C.E. Dooling is Assistant Professor of Law at The Ohio State University and Jordan Ascher is Policy Counsel at Governing for Impact.