Training without shackles: copyright, US courts, and the growing transatlantic AI divide

Training without shackles: copyright, US courts, and the growing transatlantic AI divide

Journal article

1 online resource

This article examines the first two U.S. court decisions to address whether training generative AI models on copyrighted content qualifies as fair use: Bartz v Anthropic and Kadrey v Meta Platforms, both decided in June 2025 by the Northern District of California. In reviewing the courts’ reasoning under all four statutory factors, it argues that the rulings establish a defensible legal standard, one that shields developers when training is non not intended as substitute or surrogate for the underlying work, it is transformative and does not supplant existing markets. The author offers an economic justification for this approach, showing how it lowers transaction costs, encourages innovation, and reinforces U.S. leadership in AI. The article endorses a coherent doctrine that reconciles copyright with the public interest in technological progress. It also contrasts this emerging clarity with the EU’s AI Act—a bureaucratic misadventure where innovation quietly suffocates. It concludes that U.S. judges, by declining to apply outdated legal constraints to a transformative technology, effectively acted as agents of macroeconomic policy or statecraft, preventing generative AI from being confined within a framework designed for an earlier era, and thereby reinforcing the United States’ continued leadership in technological innovation.

Crossref

European Intellectual Property Review, 0142-0461 ; E.I.P.R. 2025, 47(10), 604-613

[New York, NY] : Thomson Reuters, 2025.

English

https://1.next.westlaw.com/Copyright

External Online Journals > European Intellectual Property Review