In a groundbreaking legal decision, a federal judge in San Francisco ruled in favour of an AI company in a AI copyright law infringement lawsuit brought by a group of authors. The ruling, issued on Monday, marks the first substantial legal interpretation of how the “fair use” doctrine applies to generative AI systems. According to the court, AI companies may lawfully train their large language models on copyrighted material, as long as they acquire those materials through legal means.
This case sets a precedent that could reshape the legal landscape surrounding the use of copyrighted content in artificial intelligence. The fair use doctrine allows third-party usage of copyrighted works without the creator’s consent in specific situations, such as educational purposes or commentary. The court’s ruling affirmed that training AI models, if done with legally acquired content, falls within this exemption, much to the relief of AI developers and the concern of content creators.
Authors Challenge Anthropic’s Use of Their Works Amid AI Copyright Law Debate
The lawsuit was filed in 2024 by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who accused Anthropic AI of using AI copyright law books without permission to train its AI chatbot, Claude. The authors claimed that Anthropic had ingested millions of digitised books, some of which were their own, into its large language models. These books were either purchased and scanned or downloaded from online sources
In their complaint, the authors accused the company of bypassing proper licensing channels. “Rather than obtaining permission and paying a fair price for the creations it exploits, Anthropic pirated them,” the lawsuit alleged. The plaintiffs hoped to hold the company accountable for what they viewed as a clear violation of intellectual property rights.
Judge Declares AI Training Use as ‘Transformative’
Senior U.S. District Judge William Alsup ruled that Anthropic’s use of the plaintiffs’ works constituted fair use, calling it “exceedingly transformative”. The decision emphasised that the books were not republished or sold but rather used to train an AI system in a new and innovative way. Judge Alsup also found that scanning books legally acquired in print form for internal use did not violate AI copyright law.
However, the judge also acknowledged a grey area in Anthropic’s practices. The company had allegedly downloaded millions of copyrighted books from pirate sites to build a comprehensive digital library. While this issue remains unresolved, the court’s central finding reinforces that transformative use, even by AI, can be protected under fair use when done within legal bounds.
This decision could have sweeping implications for the tech and publishing industries alike, as courts begin to navigate the complex intersection of copyright law and artificial intelligence.
