Anthropic to Pay $1.5 Billion to Settle Landmark AI Copyright Lawsuit
Artificial intelligence company Anthropic has agreed to pay $1.5 billion to settle a class-action lawsuit brought by a group of authors who accused the company of using their books without permission to train its AI chatbot, Claude.
The settlement, revealed in a court filing on Friday, was jointly presented to U.S. District Judge William Alsup in San Francisco for approval. While the agreement was first announced in August, neither the amount nor the terms were disclosed at the time.
In their filing, the authors described the deal as historic, calling it “the largest publicly reported copyright recovery in history,” surpassing all previous copyright class-action settlements and judgments.
The lawsuit was filed last year by authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson, who alleged that Anthropic – a company backed by tech giants Amazon and Alphabet (Google’s parent company) – unlawfully copied millions of pirated books to train Claude, its generative AI system. They claimed their works were used without consent or compensation, forming part of a broader pattern of copyright infringement in the rapidly growing AI industry.
The allegations mirror dozens of similar lawsuits filed by writers, artists, and news organizations against companies including OpenAI, Microsoft, and Meta, as creators push back against the use of their intellectual property to develop AI tools.
Anthropic, like other AI developers, has argued that its use of copyrighted works qualifies as fair use, a legal doctrine allowing limited use of protected material for transformative purposes. However, in a June ruling, Judge Alsup issued a split decision: he agreed that using copyrighted books to train AI could be considered fair use, but found that Anthropic violated the authors’ rights by storing over seven million pirated books in a centralized database unrelated to AI training.
A trial had been scheduled for December to determine damages, which experts said could have reached hundreds of billions of dollars. The settlement now removes the need for that trial, though Judge Alsup must still approve the agreement.
The debate over fair use in AI training remains far from settled. Shortly after Alsup’s ruling, another federal judge handling a similar case against Meta concluded that using copyrighted material without permission to train AI systems would be unlawful “in many circumstances.”
If approved, the Anthropic settlement could set a powerful precedent for ongoing and future copyright disputes in the AI industry, potentially reshaping how companies acquire data to build and train generative AI models.
