Authors are celebrating a “historic” settlement expected to be reached soon in a class-action lawsuit over Anthropic’s AI training data.

On Tuesday, US District Judge William Alsup confirmed that Anthropic and the authors “believe they have a settlement in principle” and will file a motion for preliminary approval of the settlement by September 5.

The settlement announcement comes after Alsup certified what AI industry advocates criticized as the largest copyright class action of all time. Although the lawsuit was raised by three authors—Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber—Alsup allowed up to 7 million claimants to join based on the large number of books that Anthropic may have illegally downloaded to train its AI models.

If every author in the class filed a claim, industry advocates warned, it would “financially ruin” the entire AI industry.

It’s unclear if the class certification prompted the settlement or what terms authors agreed to, but according to court filings, the settlement terms are binding. A lawyer representing authors, Justin A. Nelson, told Ars that more details would be revealed soon, and he confirmed that the suing authors are claiming a win for possibly millions of class members.

“This historic settlement will benefit all class members,” Nelson said. “We look forward to announcing details of the settlement in the coming weeks.”

Ars could not immediately reach Anthropic for comment, but Anthropic had previously argued that the lawsuit could doom the emerging company, which was started by former OpenAI employees in 2021.