A US federal court has approved a settlement between Anthropic and a group of book authors whose works were used to train the Claude chatbot. The compensation amount is $1.5 billion — the largest payout in the history of US intellectual property litigation related to artificial intelligence.
District Judge Araceli Martínez-Olguín issued the final ruling and rejected claims from individual rights holders who considered the compensation insufficient. The conflict began in 2024 when a group of writers accused the developer of downloading pirated copies of works to train the neural network.
Key legal point: The court recognized the process of training AI on texts as fair use. However, storing over 7 million unlicensed files was deemed a direct violation of the law. This distinction has become an important precedent for the entire industry.
Authors will receive approximately $3,000 per work, with over 480,000 works on the list. The judge noted that the amount is four times the minimum payout of $750 for a standard copyright infringement. By April, rights to compensation were confirmed for 440,490 works — 91.3% of the total list.
Under the terms of the agreement, Anthropic is also required to destroy all illegal copies of the books. Some authors have opted out of the deal and continue to litigate against the company separately.
My commentary as an analyst: This is a historic moment for the AI market. The affirmation that storing pirated data constitutes a violation creates serious risks for companies that have been collecting datasets "on the fly" for years. Investors should closely monitor how Anthropic (and other players, including OpenAI) will build their compliance policies. The cost of a mistake is now measured in billions.