In a significant legal breakthrough for the AI industry, Federal Judge William Alsup has ruled that it was legal for Anthropic to train its AI models on published books without securing the authors’ permission. This is the first time a U.S. court has recognized an AI company’s claim that fair use protections can shield them from liability for using copyrighted content in model training.
The Anthropic fair use ruling could have far-reaching consequences for ongoing lawsuits brought by authors, artists, and publishers against companies like OpenAI, Meta, Google, and Midjourney. While this decision doesn’t set a binding precedent for other judges, it gives momentum to tech firms defending AI model training practices under existing copyright law.
What the Fair Use Decision Means for Creators and Tech Companies
The case—Bartz v. Anthropic—revolves around how the AI company used copyrighted books to train its large language models (LLMs). Plaintiffs argued that Anthropic used millions of books without permission, many of which were allegedly downloaded illegally from pirate websites. Yet Judge Alsup sided with the company on the key issue: the act of using copyrighted content for training purposes qualified as fair use.
Under current U.S. copyright law, fair use considers the purpose of the use (educational or commercial), whether the work is transformative, and whether the original market for the work is harmed. AI firms argue that training a model is transformative, similar to using images in a collage or text for parody. Critics, however, say this redefines “transformative” in a way that threatens the core of copyright protections.
This is the first real test of how courts will interpret fair use in the context of generative AI—a gray area untouched since the Copyright Act of 1976, well before the digital age or AI models existed.
Pirated Content Still Under Fire: Trial to Determine Damages
While the ruling favored Anthropic on fair use, the judge did not let the company off the hook entirely. Judge Alsup confirmed that a trial will still proceed regarding the alleged creation of a “central library” built from pirated books.
According to the lawsuit, Anthropic aimed to download and store “all the books in the world”—many of which were allegedly obtained from illegal online sources. Even though the court determined the training process met fair use standards, using pirated material to build that training set remains a serious legal concern.
As Judge Alsup wrote: “That Anthropic later bought a copy of a book it earlier stole off the internet will not absolve it of liability for theft but it may affect the extent of statutory damages.”
This means that even if the training itself is legal, the method by which Anthropic collected materials may result in substantial damages, especially if proven that the pirated copies were systematically used and stored.
