Anthropic settles AI book data lawsuit with authors

Anthropic has settled a lawsuit with a group of authors over its use of books to train its large language models, according to a filing with the Ninth Circuit Court of Appeals on August 26, 2025.

The case, Bartz v. Anthropic, centered on claims that Anthropic used both fiction and nonfiction works, many of which were pirated, as training data.

A lower court previously ruled that Anthropic’s use of the books qualified as fair use, but also found the company could face financial penalties due to the source of the materials.

Details of the settlement were not disclosed.

🔗 Source: TechCrunch

🧠 Food for thought

1️⃣ AI companies settle copyright cases even after winning fair use rulings

Anthropic’s decision to settle despite winning on fair use demonstrates how complex legal landscapes push companies toward resolution rather than appeals.

The court had ruled that Anthropic’s book usage qualified as fair use, yet the company still chose to settle rather than pursue its appeal to completion1.

This reflects broader industry behavior where AI companies prioritize certainty over potentially favorable but lengthy legal battles.

Even victories on core copyright issues don’t eliminate all legal risks, as companies face multiple layers of potential liability beyond the fair use determination itself.

The settlement suggests that favorable fair use rulings may not provide the comprehensive legal protection AI companies need to operate confidently.

2️⃣ Pirated training materials create separate legal vulnerabilities beyond copyright infringement

Anthropic’s case reveals how the source of training data matters as much as how it’s used, creating distinct legal challenges even when fair use is established.

Despite winning the fair use argument, Anthropic still faced “significant financial penalties for its conduct connected to the case” because many books in their training dataset were pirated1.

This distinction between fair use of content and lawful acquisition of that content creates a two-pronged legal test that AI companies must navigate.