In June 2025, a U.S. federal court in San Francisco issued a landmark ruling on the legality of training artificial intelligence (AI) models on books without the authors’ consent. The case concerned Anthropic, a company that used legally purchased books to train its AI model, Claude. The court held that such use falls within the scope of fair use and does not violate copyright law.
Background of the Anthropic Case – Using Books to Train AI
Anthropic was sued by a group of authors who alleged unauthorized use of their books to train the AI language model. The plaintiffs claimed that Anthropic not only used legally acquired copies but also millions of pirated copies downloaded from illegal sources such as Library Genesis.
Key Court Findings – Fair Use in AI Training
Judge William Alsup emphasized that:
- Training AI is a transformative use — AI learns language patterns rather than copying the book contents verbatim.
- Using books for machine learning is analogous to a reader being inspired by a book to create their own work.
- The AI model Claude incorporates safeguards to prevent reproducing protected content.
- Using legally purchased books is lawful, whereas mass downloading of pirated copies constitutes copyright infringement.
Significance of the Ruling for AI and Copyright Law
This decision sets an important precedent for the development of generative AI and clarifies the boundaries of fair use in machine learning. It demonstrates that copyright law can flexibly adapt to rapid technological advances while protecting creators’ rights. For technology companies, it signals that AI development is legally feasible if intellectual property rights are respected and pirated sources avoided.
Implications for Poland and the European Market
Although the ruling applies to U.S. law, it carries significant implications for Poland and the European Union. As AI adoption grows, issues surrounding fair use and copyright require modernized legal frameworks and regulatory adjustments to address technological challenges.