Two Stanford authors sue Anthropic and want $150,000 a book

In May 2026, the Stanford professors Tobias Wolff and Chang-rae Lee filed the lawsuit Cruz v. Anthropic. They left the class settlement of $1.5 billion in the earlier lawsuit Bartz v. Anthropic. They tell that Anthropic downloaded and kept their books from pirate websites to train Claude, its AI model. They want a jury trial and a maximum of $150,000 for each work. The source gives no reply from Anthropic.

What the lawsuit tells

The two authors tell that Anthropic downloaded, copied and shared their books from “shadow libraries”, which are websites with pirated books, and kept the files in one central library. If they win, they want a permanent injunction, which is a court rule that stops Claude training on their works.

The numbers

• The Bartz settlement went to authors and publishers of almost 500,000 registered works.
• It gave about $3,000 for each work.
• The law gives a maximum of $150,000 for each work in a full claim.
• Wolff tells that Anthropic used 10 of his publications with no permission.

What is not known

In Bartz, Judge William Alsup ruled that a permanent central library of pirated books is not legal. He also ruled that training on books that a company gets legally is fair use. The source gives no trial date. More than 70 other infringement lawsuits are against AI companies.

Insight: Anthropic guards what Claude gives out, and authors fight over what went in

In October 2026, Anthropic told that a new protection cut the copying of Claude by rival labs by about 90 percent. Two authors want a jury to put a price on the books that went into Claude. The two fights are about who can use the work of one party with no licence. It seems that the cost of training data moves from free to priced.