Skip to main content
Find a Lawyer

Artificial Intelligence Company Pays $1.5 Billion in Settlement, but Did It Really Lose?

Kit Yona, M.A.

Article by: Kit Yona, M.A.

Legal Writer

Reviewed by Joseph Fawbush, Esq. | Last updated on

No matter how you slice it, $1.5 billion is a lot of money. There are billionaires and corporations who could shrug off a payment like that and continue on without any major concerns, but those are few and far between. Despite that, artificial intelligence (AI) company Anthropic might be smiling the whole time they write out the check to a group of 500,000 authors due to a ruling earlier this summer.

In June 2025, a federal judge granted partial summary judgment to Anthropic, ruling that the company's use of legally purchased books that were then digitized and used as training material for its AI was legal under Fair Use. With dozens of other intellectual property theft cases between authors, artists, and musicians making their way through the courts, U.S. District Judge William Alsup's decision may prove to be a landmark case that sets a precedent for future rulings.

Despite its partial victory, Anthropic recently agreed to a massive settlement. Announced on September 5, the deal negates the need for a trial over works that Anthropic had downloaded from a pair of websites selling pirated copies of books. Judge Alsup declined to grant Anthropic's motion to treat pirated works the same as other works used for training Claude, Anthropic's generative AI. Judge Alsop will still need to approve the settlement.

Words, Words, Words

Anthropic was formed in 2021 by seven former employees of OpenAI, who reportedly left over issues of safety and direction. The company's primary product is the Claude AI, a Large Language Model (LLM). LLMs are AI systems designed to grow in accuracy, ability, and skill by being fed as much information as possible. Anthropic's Claude LLM features AI programs such as Sonnet, Haiku, and Opus.

Anthropic used wide-scale digitalization to increase the growth of Claude's different models. The company bought books, removed the covers and binding, and then scanned the pages in to create digital copies of the texts. These were then used to teach the AI a better understanding of how languages work, which would be harnessed to create new and original text.

In the June 2025 ruling, the judge determined that digitizing a legally obtained text was a fair use of intellectual property. Since additional copies weren't being produced and distributed to the public, Judge Alsup ruled that the change from print to digital was transformative enough to avoid copyright infringement. However, not every part of the summary judgment went Anthropic's way.

It Has "Pirate" in the Name

While the decision allowed legally obtained books to be digitized and used to train AI, it drew the line at texts that were collected by more nefarious means. It was determined that Anthropic had gained access to millions of books through sources like Library Genesis and Pirate Library Mirror, which are online repositories of pirated books.

Anthropic's executives had knowingly downloaded millions of these pirated texts, which prompted one of the suits it was facing. While Judge Alsup's ruling covered the transformation of legally obtained books, it didn't absolve the company from the copyright infringement violations of stealing works through a pirate site.

Faced with the potential of a maximum penalty of $150,000 for each transgression, Anthropic instead came to an agreement to pay $3,000 per work. It also promised to delete all of the illegally pirated texts, which it claims were never given to Claude. The potential settlement still requires Judge Alsup's approval.

Tip of the Iceberg. Bookberg. Whatever.

Whether the settlement deal is approved, there are dozens of other cases pending from creators against AI firms and their allegedly sticky fingers. Some have likened the case to what happened to Napster and the other file-sharing companies that were accused of copyright infringement over 20 years ago, but the details aren't a perfect match. The Anthropic settlement will likely influence other lawsuits, but because it never made it to trial, it didn't set an official legal precedent.

While $1.5 billion is a staggering amount of money, it's a small percentage of the investment capital that continues to flood the coffers of AI companies. Given the importance of the summary judgment supporting the transformative text defense, it may turn out to be quite the bargain.

Was this helpful?

Copied to clipboard