US District Judge Araceli Martinez-Olguin granted final approval on July 20, 2026 to Anthropic's $1.5 billion settlement with a class of authors who sued the company for using pirated copies of their books to train its Claude large language models. The case, known as Bartz v. Anthropic, is the largest copyright settlement in US history Wolters Kluwer.

The lawsuit was filed in 2024 by a group of authors who accused Anthropic of training Claude on pirated copies of their work. A large language model, or LLM, is the type of AI system that generates human-like text; it learns patterns from massive quantities of written material. The court certified the case as a class action in July 2025, and the parties reached a settlement agreement in August 2025 Wolters Kluwer. It is the first settlement reached in any of the copyright lawsuits pending against AI companies over training data Reuters.

The case was originally overseen by Judge William Alsup, who issued a split ruling on the merits. Alsup held that Anthropic's use of the authors' work to train Claude qualified as fair use, meaning the company did not need permission to use the texts for that purpose. But he found that Anthropic had infringed copyright by storing approximately 7 million pirated books in a central library. Alsup then scheduled a trial for December 1, 2025 to determine Anthropic's liability for the piracy Authors Guild.

Before that trial could take place, Alsup rejected the proposed settlement in September 2025, calling it incomplete and saying it shoved terms "down the throat of authors" Engadget. That same month, however, Martinez-Olguin, who had taken over the case after Alsup retired, granted preliminary approval on September 25, 2025 Reuters. Final approval followed on July 20, 2026 TechCrunch.

The settlement covers more than 480,000 works, with authors expected to receive approximately $3,000 per affected work. Anthropic is required to destroy the pirated copies of the books as part of the agreement. The company has launched an official settlement website at anthropiccopyrightsettlement.com, which includes a list of affected works so authors can file claims Engadget.

The settlement fund could grow beyond the initially estimated 500,000 books if additional affected works are identified Reuters. Authors' lawyers have sought $300 million in legal fees from the $1.5 billion fund Reuters.

Not all plaintiffs are on board. Some authors opted out of the settlement, finding the per-work payout too low, and have filed separate lawsuits against Anthropic Engadget.

The broader context matters for anyone building products on top of large language models. Alsup's fair-use finding on training itself remains the most consequential legal thread in this case. If that holding survives appeal, it establishes that ingesting copyrighted text for the purpose of model training can qualify as fair use, even when the source material was obtained without authorization. The infringement finding was narrowly tied to Anthropic's storage of 7 million pirated books in a central library, a factual pattern that other AI labs may or may not share depending on their data-acquisition methods.

What the settlement does not do is resolve the fair-use question across the industry. It binds only the class members in Bartz v. Anthropic. The parallel cases against OpenAI, Meta, and others are proceeding on their own tracks, and none has reached a comparable resolution. Companies that sourced training data through licensing agreements, scraped the open web, or relied on library partnerships face different factual records and will be adjudicated accordingly.

The $3,000-per-work figure has drawn criticism from opt-out plaintiffs, and the $300 million fee request from class counsel will test how courts view attorney compensation in mega-settlements involving intellectual property classes. The opt-out lawsuits will proceed separately and could produce different liability outcomes if they reach trial.

In this author's view, the Anthropic resolution offers a rough pricing benchmark for AI companies still litigating or settling copyright claims: roughly $3,000 per work at issue, scaled across nearly half a million titles. Whether that figure becomes a floor for future settlements or an outlier driven by Anthropic's specific circumstances will depend on how the remaining cases resolve. The destruction requirement for pirated copies also sets a concrete remedial precedent, one that other defendants may face in negotiations or at judgment.