Cary Coglianese

Scholarship · Administrative law

Cary Coglianese is a legal scholar whose AI work addresses two related questions: how public agencies can use algorithms lawfully, and how governments should regulate AI development and deployment. His research brings administrative-law principles and the evaluation of regulatory strategies to debates about automated decision-making and continuing human oversight.

Roles and affiliations

Coglianese is the Edward B. Shils Professor of Law and Professor of Political Science at Penn Carey Law and directs the Penn Program on Regulation. His faculty biography identifies him as a senior fellow of the Administrative Conference of the United States, with earlier service as a public member and chair of its Rulemaking Committee. He also founded and advises The Regulatory Review.

Contributions and positions

Administrative law and automated government

In Administrative Law in the Automated State (2021), Coglianese asks whether increasing reliance on machine learning is compatible with established administrative law. He argues that responsible automation could support expertise and democratic accountability rather than necessarily undermine them. The relevant comparison is with actual government decision-making, including its existing weaknesses, rather than an idealized human bureaucracy.

He nevertheless identifies a distinct concern about empathy. Even an accurate and accountable automated administration could have difficulty responding to the human circumstances of those affected. The article therefore treats the institutional design of public services as part of the legal problem, alongside predictive performance and formal procedural compliance. Its analysis is scholarship about possible arrangements, not a blanket conclusion that any agency's system is lawful.

Management-based AI regulation

With Colton R. Crum, Coglianese proposes a flexible approach in Leashes, not guardrails (2025). The authors argue that AI's varied uses and changing capabilities make fixed prescriptions an inadequate general regulatory strategy. They instead favor requirements for organizations to manage risks under continuing human responsibility.

Their proposed approach makes oversight an ongoing obligation across development, testing and deployment. Flexibility is not a call for leaving AI unregulated: regulators would require management processes while allowing adaptation to different technologies and contexts. The authors also identify implementation questions about how such processes can be made effective. This analysis connects to the U.S. policy overview and its discussion of organizational safety duties.

Selected works

• Administrative Law in the Automated State (2021) — open-access journal article.
• Leashes, not guardrails: A management-based approach to artificial intelligence risk regulation (2025), with Colton R. Crum — journal article available through the Penn Program on Regulation.

Coverage

Penn's faculty biography and bibliography document Coglianese's institutional roles and his broader work on government AI, procurement and algorithmic accountability. The bibliography distinguishes publications from working papers, which should not be treated as interchangeable publication statuses.

Related articles

• United States policy on catastrophic AI risk
• AI in Government Act of 2020
• OMB Memorandum M-25-21
• Alondra Nelson
• Frank Pasquale