Anthropic agrees to 1.5 billion dollar settlement over pirated books

A federal judge in San Francisco granted final approval on Monday to a 1.5 billion dollar settlement requiring artificial intelligence company Anthropic to compensate thousands of authors and publishers. The case, which is considered the largest known copyright recovery in history, resolves allegations that the company used pirated copies of books to train its Claude chatbot.

The litigation began in 2024 when a group of authors, including thriller novelist Andrea Bartz, initiated a class-action lawsuit. The plaintiffs alleged that Anthropic, which is backed by Amazon and Alphabet, utilized unauthorized versions of their works to teach its artificial intelligence systems to respond to human prompts. While a previous ruling by U.S. District Judge William Alsup found that training AI models on copyrighted material can constitute fair use, the court separately determined that Anthropic had infringed on copyrights by maintaining a centralized digital library containing more than seven million pirated books.

Media additions

The case was initially slated for a trial in December to determine the extent of damages related to the improper storage of these materials. Legal experts had estimated that the company’s potential financial exposure, had the case proceeded to trial, could have reached into the hundreds of billions of dollars. Instead, the parties reached an agreement that saw Judge Alsup provide preliminary approval for the settlement last September.

Following his retirement, U.S. District Judge Araceli Martínez-Olguín issued the final ruling this week. In her decision, Judge Martínez-Olguín dismissed objections from various authors who claimed the settlement was insufficient, that it improperly excluded certain copyright holders, or that the plaintiffs' legal team received excessive compensation. The court approved more than 101 million dollars in attorney fees, a figure lower than the 187.5 million dollars originally requested by the firm. Addressing the concerns regarding the settlement's total value, the judge noted that the arguments provided by objectors failed to account for the actual risks and potential outcomes of continuing the litigation through a full trial.

According to figures provided to the court, more than 91% of the over 482,000 books covered by the agreement have already been claimed by rights holders. These authors and publishers are expected to receive approximately 3,000 dollars per book.

"We are gratified by the Court’s ruling granting final approval of this historic settlement. It is the largest known copyright recovery in history. We look forward to making distributions to the Class as promptly as possible."

This resolution marks the first major conclusion among a series of active copyright lawsuits involving artificial intelligence developers and various content creators. Despite the finality of this agreement for the majority of the class, the legal status surrounding AI training data remains contested. Several authors and publishers previously chose to opt out of the class-action settlement and continue to maintain separate, ongoing lawsuits against the company regarding its AI training practices.

The court's intervention highlights the emerging complexities in legal standards for generative AI. While the judicial system has established that training models on copyrighted content may qualify as fair use, the specific infrastructure used by developers to store or cache datasets remains a point of significant litigation. The company is expected to move forward with the distribution of funds to the class as the next phase of the settlement.