Anthropic will pay $1.5 billion to authors whose pirated books it used to train its AI in the largest copyright settlement in U.S. history.

Federal Judge Araceli Martinez-Olguin in Oakland, California, granted final approval of the settlement Monday. The record payout was a modest win for copyright holders, setting a precedent as a wave of AI-related lawsuits works its way through courts.

Although the judge said Anthropic could use authors' works for training, she backed the payout for acquiring and storing pirated books.

More than 500,000 authors and publishers were part of the class-action lawsuit, and they will get paid an estimated $3,000 per eligible work. Their lawyers were awarded $101.6 million in fees.

In 2024, a group of authors sued Anthropic, arguing that the company pirated hundreds of thousands of copyrighted books and used them without permission.

U.S. District Judge William Alsup, who initially pronounced his judgment in June last year, offered a partial victory for the authors.

The court found the use of books to train AI was allowed under fair-use norms, but the manner in which Anthropic acquired them — downloading millions of books from pirated libraries without permission — was not.

The problem was not that Anthropic had used books to train its model, but rather that it used them without permission or regular payment.

"Two years after we filed, the settlement for our class-action lawsuit got final approval," Andrea Bartz, best-selling author and one of the lead plaintiffs in the lawsuit, said in a post on Instagram. "As I've been saying from the start, this is an important first step toward accountability for Big AI's breathtaking theft."

She added: "I'm glad authors and publishers could tell Anthropic the obvious: You can't steal our stuff!"

This lawsuit is one of many filed by copyright owners over alleged unauthorized use of their work to train large language models.

Some 350 authors in the Anthropic lawsuit opted out of the settlement, with some expressing objections and betting they can recover significantly more through individual litigation.

What constitutes "fair use" of copyrighted works for training an AI model remains contentious. While wholesale plagiarism and the copying of large sections of someone's content without permission or acknowledgment of the original source is not allowed, learning from and citing others as well as sharing samples of their work is often considered fair use.

What AI does with words it devours is different from just copying and regurgitating the content verbatim, so many people consider it fair use.

The technology underlying chatbots such as Claude and ChatGPT uses patterns in data by analyzing vast quantities of digital text and video from the internet. Tech companies extract information from books, news articles and Wikipedia pages, which copyright holders said is often done without consent, compensation or control over how it is used.

In U.S. copyright law, the benchmark for using a copyrighted work without permission is to examine whether it is transformative, adding significant extra value to the output, rather than merely copying the work.

"The use of the books at issue to train Claude and its precursors was exceedingly transformative and was a fair use," Alsup noted in his earlier judgment.

Anthropic also purchased millions of print copies in bulk, stripped the books' bindings and made scanned digital copies that were used to train Claude. This print-to-digital format change was also fair use, the judge ruled.

"The fight over whether AI training is fair use is yesterday's war," James Rubinowitz, an adjunct professor at Cardozo School of Law, said in a statement. The Anthropic judgment ruled that training on lawfully acquired books is "quintessentially transformative."

"The live question in 2026 is acquisition: where the books came from, who paid for them, and what the company knew about their provenance when it hit download," he said.

The debate will continue in the courts, which are also looking at other copyrighted works such as news articles and movies.

The New York Times has a lawsuit against OpenAI and Microsoft for training on copyrighted news pieces, and Disney and Universal have sued image generation company Midjourney for training on allegedly unauthorized content and creating reproductions of famous characters.

©2026 Los Angeles Times. Visit at latimes.com. Distributed by Tribune Content Agency, LLC.