Court Ruling Splits Decision on Anthropic: Legal Book Use Deemed Fair, Pirated Copies a Clear Copyright Violation
In a major decision for the evolving landscape of AI and copyright law, U.S. District Judge William Alsup delivered a split ruling in the copyright infringement case against AI company Anthropic, known for developing the Claude chatbot. The case, brought by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, alleged that Anthropic unlawfully used their copyrighted works to train its AI models.
While the court found that Anthropic’s use of lawfully acquired books qualified as fair use, Judge Alsup drew a firm legal line against the company’s use of pirated content — a move that could have far-reaching consequences for the AI industry.
Transformative Use vs. Pirated Inputs
At the heart of the case were two very different methods Anthropic used to source training materials. First, the company spent millions to acquire and digitize print books, including used copies, which it then used to train Claude and its earlier iterations. On this front, Judge Alsup ruled decisively in Anthropic’s favor, stating the company’s use of the materials was “exceedingly transformative” and aligned with the fair use doctrine under Section 107 of the Copyright Act.
“Anthropic spent many millions of dollars to purchase millions of print books,” the judge wrote. “It acquired copies of millions of books, including of all works at issue for all plaintiffs.” The court emphasized that the AI model Claude did not memorize or replicate substantial portions of copyrighted texts, nor did it output infringing content. The authors’ complaint centered solely on inputs, not outputs — a detail that set this case apart from similar lawsuits currently pending against companies like OpenAI and Microsoft.
Pirated Books: A “Blatantly Infringing” Practice
However, the court’s tone shifted sharply when addressing Anthropic’s use of pirated works. The judge made clear that the company’s download of over 7 million illegally sourced book files — despite the fact that it ultimately decided not to train on them — was an irredeemable violation of copyright law.
“Anthropic downloaded over seven million pirated copies of books, paid nothing, and kept these pirated copies in its library even after deciding it would not use them to train its AI,” Judge Alsup wrote. “Authors argue Anthropic should have paid for these pirated library copies. This order agrees.”
Further condemning the action, the judge dismissed any potential argument that this was an innocent or inconsequential act. “Such piracy of otherwise available copies is inherently, irredeemably infringing even if the pirated copies are immediately used for the transformative use and immediately discarded,” he wrote.
No Copyright Loopholes for AI Developers
In contrast to Google’s legally meticulous handling of book digitization for the Google Books project — which involved legal contracts and secure technical controls — Anthropic’s approach was found lacking. The ruling highlights deposition testimony in which CEO Dario Amodei reportedly described the decision to use pirated books as a way to avoid the “legal/practice/business slog.” Even after changing their mind about training on the pirated books “for legal reasons,” the company held onto them.
“Anthropic seems to believe that because some of the works it copied were sometimes used in training LLMs, Anthropic was entitled to take for free all the works in the world and keep them forever with no further accounting,” Alsup wrote. “There is no carveout, however, from the Copyright Act for AI companies.”
What Happens Next?
The fair use ruling for lawfully acquired books ends one portion of the lawsuit, but the case will proceed to trial to assess the damages related to Anthropic’s use of pirated works. The authors may seek either actual damages or statutory damages for what Judge Alsup called “willful” infringement — potentially exposing the company to steep financial penalties.
While this ruling doesn’t apply to other high-profile AI copyright suits, it may influence future decisions by drawing a clearer distinction between transformative fair use and blatant infringement. For the publishing and literary world, it underscores a critical legal stance: buying the books matters — and stealing them, even for AI training, is still theft.