The case against Anthropic began in 2024, when three authors claimed that the company had pirated their work to train the chatbot, Claude.
In June 2025, a federal judge ruled that it was legal to train Claude with copyrighted books, but that it was illegal to pirate those books, which is what the company had done. In September 2025, that same judge granted preliminary approval to the settlement which was officially approved on Tuesday.
This marks the first major settlement out of dozens of AI copyright lawsuits being filed in the U.S. today. The authors’ lead lawyer, Justin Nelson, said it was “the largest known copyright recovery in history”.
Brittany Newell is San Francisco based author whose 2017 book Oola, was pirated to train Claude. She said she was surprised when she found her book in the lawsuit’s database.
“I guess I just assumed they would be using, like, the great books or, I don't know, Colleen Hoover, popular books. And my first book, Oola, was not popular.”
Newell says it enrages her to know that AI companies are stealing from artists, but that she’s soothed by the fact that most of it is still “AI Slop.”
“Every AI trained piece of writing that I've encountered has been like so bad, and so obvious, so I try to counteract the sense of injustice with like, wow, well, they did all that and in my opinion it's still a huge failure.”
For the author, her frustration is less about her book being pirated, than it is about AI in general.
“It extends outwards to the rage I feel for, like, what AI has done to san francisco, what it’s doing to the environment, the sort of crazed, gold rush attitude.”
More than 480,000 books were covered by the ruling. About 91% have been claimed by authors or publishers who are now waiting to get paid. Anthropic will pay about $3000 per book.
