Anthropic used books by Stanford professors, including Tobias Wolff and Chang-rae Lee, to train Claude after obtaining copies from online “shadow libraries,” according to a lawsuit described by The Stanford Daily. In May 2026, a federal judge ruled that Anthropic could not pirate books to create a permanent library, while saying training on lawfully acquired books could qualify as fair use.
The ruling came in Bartz v. Anthropic, which led to a $1.5 billion settlement covering nearly 500,000 registered works. Wolff and Lee opted out of that settlement and, with other authors, filed a separate lawsuit, Cruz v. Anthropic, in May 2026. They allege Anthropic illegally downloaded, copied and stored their books. Those claims are allegations, not findings in the new case.
What did the judge allow?
Judge William Alsup drew a line between how Anthropic obtained books and how it used them. He compared training on lawfully acquired books to a person reading and remembering a work’s themes and style, and said that kind of training could be fair use. But he barred Anthropic from pirating books to build a permanent central library.
That distinction matters because the settlement did not resolve every author’s claims. The authors in Cruz are seeking a jury trial and statutory damages of up to $150,000 per infringed work. The Stanford Daily reports that the Bartz settlement offered about $3,000 per work. If the authors prevail, they also want an order permanently stopping Claude training on their books.
A wider fight over training data
Wolff said 10 of his publications were used to train AI without permission. He says the settlement still leaves Anthropic broad freedom to draw on literature, while Stanford Law scholar Paul Goldstein points to a practical obstacle: a training run can involve millions of copyrighted works, making individual licensing difficult.
The dispute is part of a wider legal contest. The Stanford Daily reports more than 70 copyright lawsuits against AI companies, including cases brought by The New York Times and Getty Images. The Authors Guild is also pursuing a case against OpenAI and Microsoft and working with Congress on legislation that would require licensing. The legal rules for large-scale AI training remain unsettled.
Why it matters
For authors whose books may be used in AI training, the Bartz ruling distinguishes between training on lawfully acquired books and building a library from pirated copies. The authors in Cruz are asking for larger damages and a ban on training Claude on their work, but the case’s outcome—and the broader rules for AI training—remain unknown.
