Anthropic settles AI learning copyright infringement lawsuit with author group, avoiding up to $1.5 trillion in damages

AI development company Anthropic and a group of authors have reached a settlement in a class action lawsuit over copyright infringement related to AI model training data. Court documents reveal that the parties signed a letter of intent dated August 25, 2025, outlining the terms of the proposed settlement.

Anthropic Settlement | DocumentCloud
https://www.documentcloud.org/documents/26075972-anthropic-settlement/

Authors celebrate “historic” settlement coming soon in Anthropic class action - Ars Technica
https://arstechnica.com/tech-policy/2025/08/authors-celebrate-historic-settlement-coming-soon-in-anthropic-class-action/

The lawsuit was filed by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson against Anthropic, alleging that the company infringed their copyrights by improperly using a dataset containing pirated books to train its AI model, 'Claude.'

Anthropic sued by three authors for copyright infringement, claiming it used hundreds of thousands of copyrighted books in training for Claude - GIGAZINE

In June 2025, the U.S. District Court for the Northern District of California ruled that training an AI model with legally purchased books constitutes fair use. However, the court also acknowledged that the data from pirated websites is itself illegal and infringes copyright, and a separate lawsuit will be filed to determine damages for Anthropic's use of the pirated data.

A ruling states that AI companies don't need permission from the author to train their AI with legally acquired books.

If the damages were determined, Anthropic could be sued for more than $1 trillion (approximately 150 trillion yen), which could financially ruin the entire AI industry depending on the outcome of the trial.

Edward Lee, a professor of law at Santa Clara University, called Anthropic's decision to settle after a lengthy legal battle 'a surprising development,' suggesting the company was forced to settle to avoid the risk of facing potentially devastating statutory damages.

While the specific terms of the settlement have not yet been made public, plaintiffs' attorney Justin Nelson said, 'This is a historic settlement that benefits all class members.' The settlement is expected to be finalized on September 3, 2025, and a motion for preliminary approval is expected to be filed with the court by September 5.

Ars Technica, an IT news site, analyzed that while the settlement lost an opportunity to establish a legal precedent because a June 2025 court ruling found the use of copyrighted material in AI training to be 'fair use,' the basic judicial ruling on fair use remains in place. It also expressed the view that this settlement will prevent the entire AI industry from being embroiled in countless lawsuits.