75 Am. U. L. Rev. F. 137 (2026).

Abstract

Generative AI systems are trained on vast quantities of data, yet the use of copyrighted materials in training often remains opaque to both consumers and creators. This opacity produces dual harms—consumers cannot meaningfully evaluate the products and services they use, and creators lack the information necessary to detect the use of their works and to enforce their statutory rights under copyright law. While the legality of training on unlicensed copyrighted works remains unsettled under the copyright fair use doctrine, this Note argues that the Federal Trade Commission need not wait for judicial resolution to address the harms resulting from unlicensed use. This Note further contends that Congress should provide a clear foundation for FTC action, whether by declaring unlicensed use of copyrighted training data unlawful or by expressly requiring disclosure of unlicensed use, to reduce the agency’s institutional hesitancy to act. Such Congressional action would address current information asymmetries between creators and AI developers, support emerging licensing markets, and give the FTC a clear mandate to enforce meaningful transparency obligations.

* Senior Technical Editor, American University Law Review, Volume 75; J.D. Candidate, May 2026, American University Washington College of Law; M.A. Candidate, International Affairs, Dec. 2026, American University School of International Service; B.A., Chinese Language and Culture, 2017, University of Puget Sound. I would like to thank Tom Clees for introducing me to copyright policy and inspiring this Note. I am also grateful to the American University Law Review staff for being such a supportive and collaborative team, and to my family for their constant support and encouragement.