Anthropic trained Claude on illegally obtained Stanford professors' books

What it claims

The Stanford Daily reports that books by more than 10 Stanford professors were among the pirated texts Anthropic used to train Claude. Named authors include Tobias Wolff, Abraham Verghese, Chang-Rae Lee, Daniel Mason, Hayagreeva Rao and Myra Strober. Ten of Wolff's publications were used without permission.

The page names the sources at issue. The plaintiffs allege that Anthropic downloaded, reproduced and distributed copyrighted books from the shadow libraries LibGen and PiLiMi. They also allege that storing those files in a permanent central library was itself unlawful. This is the conduct Judge William Alsup ruled against in Bartz v. Anthropic. Training on lawfully acquired books was held to be fair use. Alsup's reasoning was that the models train on works "not to race ahead and replicate or supplant them — but to turn a hard corner and create something different."

Cruz v. Anthropic, filed in May 2026, involves Wolff, Lee and 27 other authors who say the settlement is not enough. It is a non-class suit seeking a jury trial. Beyond statutory damages, a win would give the plaintiffs a permanent injunction against training Claude models on their works. Wolff said the authors' agents judged that the class action "didn't think the blanket lawsuit covered quite enough of the liability."

The wider positions on licensing:

• Authors Guild. CEO Mary Rasenberger says licensing should let rights holders require output filters, so that their works do not appear in model outputs unless separately licensed. The Guild is suing OpenAI and Microsoft. It is also working with Congress on national legislation that would require licensing.
• Paul Goldstein. Goldstein, a Stanford Law IP scholar and a Bartz class plaintiff, takes the opposite view of feasibility. He says licensing every work, "or, indeed, any," is impracticable because of transaction costs. He also notes that the U.S. alone among industrialized nations lacks a moral right of attribution.
• Other suits. More than 70 infringement suits have been filed against AI companies. Plaintiffs include The New York Times and Getty Images.

Limits

This is a student news report, not a court document. It quotes no filing text and gives no hearing dates or status for Cruz. The opening sentence attributes a judge's ruling to "a May 2026 lawsuit." Elsewhere the page places the Alsup ruling in Bartz and the May 2026 filing in Cruz, so the dates should be checked against the dockets.

The page states that the legal status of mass scraping for LLM training, outside the pirated-books question, remains unsettled.