A federal judge in San Francisco has granted final approval to Anthropic’s $1.5 billion settlement with authors and publishers who accused the artificial-intelligence company of illegally downloading copyrighted books and using them in the development of its Claude chatbot. The agreement represents the largest known copyright settlement in United States history and marks a significant victory for creators challenging how technology companies obtain material for AI training.
The class-action lawsuit was filed in 2024 by authors Andrea Bartz, Charles Graeber and Kirk Wallace Johnson. They alleged that Anthropic collected millions of books from online piracy repositories and stored them in a central digital library. The settlement covers approximately 482,000 eligible books, with authors and publishers expected to receive roughly $3,000 for each work before deductions for legal fees, administrative expenses and other costs. More than 91% of eligible works had already been claimed when the agreement received final approval.
The dispute produced an important legal distinction between using copyrighted works to train AI and obtaining those works through piracy. In an earlier ruling, a federal judge concluded that using books to train Anthropic’s models could qualify as fair use because the process was transformative. However, the judge found that downloading and retaining pirated copies created a separate copyright problem that could expose the company to substantial damages.
Anthropic faced the possibility of a trial involving potentially enormous financial liability. Rather than continue litigating, the company agreed in 2025 to establish the $1.5 billion compensation fund without admitting wrongdoing. The settlement also resolved claims involving books allegedly taken from websites and databases that distributed copyrighted material without authorization.
U.S. District Judge Araceli Martínez-Olguín described the agreement as providing meaningful relief to the affected authors and publishers. She rejected objections arguing that the settlement was inadequate and determined that the recovery was reasonable considering the risks, complexity and uncertainty of continuing the lawsuit. Some rights holders had argued that approximately $3,000 per work was insufficient, particularly when statutory copyright damages can be considerably higher.
The judge also reduced the amount awarded to the plaintiffs’ lawyers. Attorneys had initially sought compensation of up to $300 million and later requested approximately $187.5 million. Martínez-Olguín approved about $101.6 million in legal fees, equivalent to roughly 6.8% of the settlement, after concluding that a larger percentage-based award would provide an excessive financial windfall.
Although the settlement concludes the class action for participating authors and publishers, it does not resolve the broader conflict over copyrighted content and generative AI. Some writers and publishing companies excluded from the agreement—or dissatisfied with its terms—are continuing separate lawsuits against Anthropic. Other technology companies also face legal claims alleging that books, journalism, music, photographs and artwork were used without authorization to develop AI systems.
The ruling may influence future negotiations between AI developers and creative industries, but it does not establish that all AI training involving copyrighted material is illegal. Instead, the case emphasizes that courts may treat the purpose of AI training differently from the methods used to acquire training data. Legally purchased or licensed works may receive fair-use protection in some circumstances, while material obtained from piracy websites can create separate liability.
The agreement delivers major compensation to authors while sending a warning to AI companies about the risks of building large training libraries from unlawfully acquired material. The decision is likely to strengthen demands for licensing agreements, transparent datasets and direct payment to creators whose work contributes to the development of commercial AI products.
