Develop-Fair Use for Artificial Intelligence: A Sino-U.S. Copyright Law Comparison Based on the Ultraman, Bartz v. Anthropic, and Kadrey v. Meta Cases

Fair use rules need updating for AI: Traditional copyright fair use frameworks were designed for human creators and cannot adequately address the unique challenges posed by generative AI systems that train on massive amounts of copyrighted material, requiring courts to develop new analytical approaches rather than simply applying old rules.

Market competition, not just legal categories, should guide AI fair use decisions: Rather than relying on rigid legal classifications, courts should focus on substantive analysis of how AI training affects different industries—examining whether AI companies are competing with or complementing publishers, and whether new licensing markets (like "training licenses") are emerging as alternatives to fair use claims.

Both U.S. and Chinese courts are moving toward flexible, case-by-case analysis: The Ultraman case in China and the Bartz v. Anthropic and Kadrey v. Meta cases in the U.S. demonstrate that courts across different legal systems are shifting from fixed rules to dynamic balancing that considers specific market conditions and competitive impacts in each situation.

The publishing industry is developing new business models to protect copyrighted works: Publishers are creating licensing markets for AI training data as a market-based alternative to fair use, which courts must consider when determining fair use boundaries—meaning the legal answer depends partly on whether viable licensing options exist.

Fair use boundaries will continue evolving with AI industry development: The paper suggests that copyright fair use for AI should not be permanently fixed but should adapt as the AI industry matures and market conditions change, requiring ongoing judicial reassessment rather than one-time legal resolution.