Anthropic Pays $1.5B to Settle Copyright Claims Over AI Training

AI startup Anthropic has agreed to pay $1.5 billion to settle copyright infringement claims from thousands of authors and publishers whose work was used without permission to train its Claude chatbot systems. The settlement, approved by US District Judge Araceli Martínez-Olguín this week, represents what attorneys describe as the largest known copyright recovery in history.

Bloomsbury, the London-based publisher behind Harry Potter and bestselling authors including Sarah J. Maas and Susanna Clarke, disclosed it has 14,087 titles covered under the agreement. The company expects to receive approximately $19 million after legal fees, with proceeds split between the publisher and affected authors. Individual titles will receive roughly $3,000 each in compensation.

Why it matters

This settlement establishes the first major financial precedent for how AI companies must account for using copyrighted material in training data. With dozens of similar lawsuits still pending against other AI firms, the outcome signals that tech companies may face substantial costs for past use of protected works—and will likely need to negotiate licensing deals going forward. The case directly challenges the industry's reliance on "fair use" defenses and could reshape how generative AI systems are developed.

The Legal Challenge

Novelist Andrea Bartz and two co-plaintiffs filed the original lawsuit in 2024, arguing that Anthropic scraped vast amounts of copyrighted content from the open web to train its AI models without seeking permission or providing compensation. The case covered approximately 482,000 works, with 91% of those titles claimed by rights holders under the settlement terms.

US-based AI companies have consistently argued that training models on publicly available data falls under fair use doctrine, which permits limited use of copyrighted material without owner consent in specific circumstances. Authors and creative professionals worldwide have rejected this interpretation, demanding either advance permission or payment for their work.

Lead attorney Justin Nelson characterized the settlement as providing "meaningful relief" to affected parties. Anthropic's deputy general counsel Aparna Sridhar said the company was pleased with the high claim rate and looked forward to closing the matter.

Shifting Business Models

Bloomsbury has separately pursued AI licensing arrangements, announcing a deal last year that allows the company to sell academic works for training generative AI systems. Authors can opt into this program and receive royalties if they choose to license their content. The publisher recently indicated it plans to expand the initiative to additional subject areas.

The dual approach—pursuing compensation for past unauthorized use while establishing forward-looking licensing frameworks—may become a template for other publishers navigating the AI landscape. Dozens of copyright lawsuits filed by authors and news organizations against AI companies remain active in US courts.

Details of the settlement were first reported by The Guardian.