Kevin Afiesh
Volume 77, Issue 3, 689-726
The legal landscape for generative artificial intelligence (AI) and copyright is now shaped more by private settlements that overwhelmingly favor copyright holders than by courts or legislators. As AI companies rely on copyrighted materials to train their models, content owners have filed lawsuits challenging these practices. However, instead of clarifying fair use limits in court, most disputes are resolved through private settlements. These agreements impose licensing fees, content restrictions, and compliance requirements, creating a shadow regulatory system that governs how AI companies access and use training data. Although courts and agencies have begun to weigh in, their interventions offer only partial guidance, highlighting the persistent uncertainty surrounding fair use in generative AI. This Note argues that settlements are not merely resolving disputes but actively reshaping AI copyright law by establishing an informal licensing system that significantly benefits copyright holders and dominant AI developers. In turn, this sidelines smaller AI companies and reduces competition. The result is a privatized framework where copyright holders dictate the terms of AI development while avoiding judicial and legislative oversight. This Note examines the motivations behind these settlements, the legal norms they create, and whether policymakers should intervene. It considers three potential paths forward: (1) Congress could codify AI training as fair use to protect innovation and disrupt the emerging pay-to-train model; (2) lawmakers could require transparency and opt-out mechanisms to balance copyright control with generative AI development; or (3) Congress could take no action, allowing private deals to continue shaping AI copyright law without intervention. The future of AI copyright law hinges on whether policymakers rebalance the regulatory environment or allow private agreements to prevail.