Global copyright laws converge to support AI training, new study reveals

Research uncovers surprising harmony in copyright exceptions for AI across diverse legal systems, predicting further convergence despite ongoing challenges in this rapidly evolving field.

Global copyright laws converge to support AI training, new study reveals
The Globalization of Copyright Exceptions for AI Training

A new study published in the Emory Law Journal reveals a surprising convergence in global copyright laws regarding the use of copyrighted works for artificial intelligence (AI) training. The research, conducted by legal scholars Matthew Sag and Peter K. Yu, suggests that countries worldwide are increasingly embracing copyright exceptions to facilitate AI development, despite varying legal traditions and economic conditions.

The study, titled The Globalization of Copyright Exceptions for AI Training, examines copyright law developments across multiple jurisdictions, including the United States, European Union, Japan, and China. It identifies an emerging international equilibrium where countries are finding ways to balance copyright protection with the demands of AI innovation.

The researchers highlight key differences in approaches among major jurisdictions:

United States: Relies on its fair use doctrine, which has been interpreted by courts to allow for nonexpressive uses of copyrighted works, including AI training.

European Union: Has introduced specific exceptions for text and data mining (TDM) in its Copyright Directive, with separate provisions for research organizations and commercial entities.

Japan: Pioneered an express exception for TDM and computational data analysis, which has been expanded to cover any use that does not result in the enjoyment of a copyrighted work.

China: Has not yet introduced specific exceptions for AI training but has amended its Copyright Law to potentially allow for such exceptions through administrative regulations.

Despite these differences, the study finds a convergence in the overall trend of facilitating AI training through copyright law. The researchers attribute this convergence to three key factors: the centrality of the idea-expression distinction in copyright law, global competition in AI development, and what they term a "race to the middle" as countries seek balanced approaches.

The authors note several uncertainties that could disrupt the emerging equilibrium, including ongoing copyright litigation in the United States, new licensing deals between AI companies and content owners, and regulatory efforts such as the European Union's AI Act.

The study concludes by emphasizing the importance of understanding the relative strengths and weaknesses of different copyright approaches in facilitating AI development. As countries continue to refine their legal frameworks, the design of copyright systems could play a crucial role in shaping the future of AI technology.

This research provides valuable insights for policymakers, legal practitioners, and AI developers navigating the complex intersection of copyright law and artificial intelligence in an increasingly globalized technological landscape.