Anthropic will pay $1.5 billion to train its AI with pirated books.

Anthropic , known for its artificial intelligence chatbot Claude , has agreed to pay at least $1.5 billion to settle a class-action lawsuit in the United States. The dispute arose because the company used books without the authors' permission to train its AI, according to court documents filed Friday.

Plaintiffs' attorney Justin Nelson called the settlement "historic," noting that it surpasses any previous copyright recovery in the AI ​​era . The ruling sets an important precedent for using copyrighted works to train artificial intelligence models .

The lawsuit was filed by writers Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson , who claimed that Anthropic had illegally copied their books to develop Claude, ChatGPT's direct competitor.

According to the authors, the company's practice not only violated their rights, but also involved the mass downloading of pirated books , creating an unauthorized digital library that was used to train artificial intelligence.

In June, Anthropic won a partial victory . San Francisco federal judge William Alsup ruled that training the AI ​​models on the books, whether purchased or pirated, significantly transformed the works , constituting “fair use” under U.S. copyright law.

The judge explained in his 32-page ruling that the transformative use of books to train AI did not violate copyright , as long as the nature of the content and its educational or research purposes were respected.

However, the judge also ruled that downloading millions of pirated books to create Anthropic's digital library did constitute infringement . This led the company to opt for a financial settlement, avoiding protracted litigation.

The settlement will cover approximately 500,000 books , with an estimated payment of $3,000 per work , which is four times the minimum legal compensation under U.S. copyright law.

As part of the agreement, Anthropic will destroy the original pirated files and any derivative copies. However, the company will retain rights to the books it legally acquired and scanned, thus protecting its ability to continue training AI models with legitimate materials.

Anthropic staff attorney Aparna Sridhar said the company remains committed to developing safe AI systems that help people and organizations expand capabilities, drive scientific discovery, and solve complex problems .

Mary Rasenberger, executive director of the Authors Guild , stressed that the agreement sends a strong message to the AI ​​industry : there are legal consequences when authors' works are used without permission .

The case highlights the need for regulation and ethics in the training of artificial intelligence models , especially when using copyrighted works.

Although a settlement has been reached, it still needs to be approved by Judge Alsup . A hearing is scheduled for next Monday in federal court in San Francisco , where the resolution will be formalized and the final terms of the agreement confirmed.

The Anthropic case represents a turning point in the relationship between copyright and artificial intelligence . The $1.5 billion payment, the destruction of pirated files, and the strengthening of the legal use of works underscore the importance of respecting intellectual property in the digital age.

This landmark agreement sets a precedent that will likely influence how AI companies train their models, ensuring that creators receive recognition and compensation for their work as the industry moves toward more ethical and regulated practices.