Global Perspectives on Fair Use and AI Training Data

Abstract

The increasing adoption of artificial intelligence (AI) has recently posed important questions regarding the propriety of copyrighted data in training datasets. Under fair use and similar exceptions and limitations as legal doctrines, it is important to evaluate the permissions for such usage. Despite this, its meaning and implementation differ greatly from one jurisdiction to another, raising legal ambiguity when developers engage in AI businesses across geographical borders. The article addresses a pressing issue. It offers a comparative examination of fair use and analogous doctrines in the United States, the European Union, and emerging markets, set against a rapidly evolving legal landscape. Recent landmark litigation in the United States involving AI developers, alongside the adoption of the EU AI Act, has brought renewed urgency and clarity to debates on the lawful use of copyrighted material in the context of AI development. This study intends to show how these frameworks can be used to explain questions concerning copyright, transformative use, and public benefit. Finally, this study examines regulatory trends and concerns about AI from a legal and policy perspective, assesses the alignment and misalignment of regulatory frameworks, and explores the challenges and possible solutions.