A U.S. federal judge has handed artificial intelligence startup Anthropic a decisive legal win, ruling that the company’s use of copyrighted books to train its AI model, Claude, falls under fair use and is “quintessentially transformative.”
The judgment, issued Monday by Judge William Alsup of the Northern District of California, is being hailed as a pivotal moment in the growing legal clash between content creators and AI developers.
The decision not only clears a major legal hurdle for Anthropic—backed by Amazon—but also provides crucial legal clarity for the broader generative AI industry, where similar lawsuits are stacking up against companies like OpenAI, Meta, and Google. It may serve as a precedent in determining whether using copyrighted material for machine learning constitutes infringement or is protected by law.
Training AI is ‘Like Any Reader Aspiring to Be a Writer’
“The purpose and character of using copyrighted works to train LLMs to generate new text was quintessentially transformative,” Judge Alsup wrote in his ruling. “Like any reader aspiring to be a writer.”
In his opinion, Alsup noted that the plaintiffs—authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson—failed to show that Anthropic’s Claude model directly reproduced their work’s “creative elements,” or mimicked any of their “identifiable expressive style.”
This, he said, means the large language models (LLMs) do not offer “market substitutes” for the original books and are not infringing on the authors’ rights. Alsup emphasized that using books to train AI systems to generate new and original outputs fits squarely within the doctrine of fair use.
Significance for Other AI Lawsuits, Including OpenAI vs. New York Times
The implications of the ruling stretch far beyond Anthropic.
OpenAI, for instance, is currently embroiled in a high-profile lawsuit filed by The New York Times, which claims its articles were used without permission to train ChatGPT. The media outlet alleges copyright violations and demands compensation or licensing agreements.
Judge Alsup’s decision is likely to be cited as a defense model in that case and others like it. AI developers will likely argue that if Claude’s training on copyrighted books is transformative and covered by fair use, then similar practices used by OpenAI and others should also be considered lawful.
Plaintiffs Claim “Massive Theft” of Books
The lawsuit against Anthropic was filed last August and accused the company of building a “multibillion-dollar business by stealing hundreds of thousands of copyrighted books.” The authors cited the inclusion of their work in a cache of roughly 7 million books that they say Anthropic used to create a centralized training dataset.
Although Alsup upheld Anthropic’s defense regarding model training, he left one critical issue unresolved. A separate trial will be held to determine whether the company improperly retained and used pirated books to build that dataset—and if so, whether damages should be awarded.
“That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for the theft,” Alsup wrote. “But it may affect the extent of statutory damages.”
Anthropic Welcomes Ruling as a Win for Innovation
Anthropic praised the court’s decision, calling it a validation of its practices.
“We are pleased with the ruling, which is consistent with copyright’s purpose in enabling creativity and fostering scientific progress,” a spokesperson said in a statement.
The company, founded by former OpenAI researchers, is among a handful of startups vying for dominance in the next wave of AI development. Its Claude model has been positioned as a rival to OpenAI’s ChatGPT and Google’s Gemini.
As AI Copyright Battles Intensify
There is a flood of copyright-related lawsuits against AI companies. In addition to the New York Times case, comedian Sarah Silverman and other authors have sued OpenAI and Meta for using their written works. Getty Images and various record labels are also seeking legal remedies for what they claim is unauthorized use of their intellectual property in training AI systems.
But the Alsup ruling could shift the momentum.
Some legal experts note that while it doesn’t settle all disputes over AI and copyright, it creates a clearer framework: If a model is not outputting recognizable or plagiarized content, and its use of copyrighted material is to build something fundamentally new, courts may view that as fair use.