TLDR: AI firm Anthropic has reached a monumental $1.5 billion copyright settlement with a class of authors and publishers. The agreement, preliminarily approved by a federal judge, resolves claims that Anthropic illegally used nearly half a million pirated books to train its Claude AI models. This landmark deal, the largest U.S. copyright settlement of its kind, signals a significant shift in accountability for AI companies regarding intellectual property.
San Francisco, CA – September 27, 2025 – In a development poised to reshape the landscape of artificial intelligence and intellectual property, AI powerhouse Anthropic has secured preliminary judicial approval for a staggering $1.5 billion copyright settlement. This historic agreement addresses a class-action lawsuit brought by authors and publishers, who alleged that Anthropic unlawfully utilized their copyrighted works, sourced from notorious ‘shadow libraries’ like LibGen and Pirate Library Mirror, to train its advanced Claude AI models.
The settlement, which received preliminary endorsement from U.S. District Judge William Alsup of the Northern District of California on September 25, 2025, marks the largest U.S. copyright settlement in history. It mandates Anthropic to pay $1.5 billion into a settlement fund, from which qualifying rightsholders will receive approximately $3,000 per book, to be split between authors and publishers. The agreement covers an estimated 465,000 to 500,000 books that were allegedly pirated.
The lawsuit, initiated in August 2024 by nonfiction authors Andrea Bartz, Charles Graeber, and Kirk Johnson, accused Anthropic of copying their books without permission. While Judge Alsup had ruled in June 2025 that training Claude on legally acquired books constituted ‘transformative’ fair use, he refused to excuse the use of pirated materials, setting the stage for this substantial settlement.
Under the terms of the deal, Anthropic has also committed to destroying the original pirated files and any copies made from them. Crucially, the release from liability is explicitly limited to past conduct, specifically Anthropic’s prior downloading and internal training activities through August 25, 2025, on the identified works. This means the company retains rights to books it legitimately purchased and scanned.
The settlement has been met with strong reactions from both sides. Mary Rasenberger, CEO of the Authors Guild, hailed the settlement as ‘a vital step’ and issued a stern warning to the AI industry: ‘AI companies cannot simply steal authors’ creative work…just because they need books to develop quality LLMs.’ She emphasized that the deal ‘sends a clear signal to AI companies that infringement of authors’ rights comes at a steep price.’ Maria A. Pallante, President and CEO of the Association of American Publishers, echoed this sentiment, calling it ‘a major step in the right direction in holding AI developers accountable for reckless and unabashed infringement.’ She added that the settlement ‘sends a powerful message to AI companies and creators alike that taking copyrighted works from these pirate websites is wrong.’
Anthropic, valued at $183 billion in September 2025, expressed satisfaction with the preliminary approval. Aparna Sridhar, Deputy General Counsel at Anthropic, stated, ‘We are pleased the court has granted preliminary approval of the settlement. The decision will allow us to focus on developing safe AI systems that help people and organizations extend their capabilities, advance scientific discovery, and solve complex problems.’ She also reiterated that ‘the court’s landmark June ruling that AI training constitutes transformative fair use remains intact. This settlement simply resolves narrow claims about how certain materials were obtained.’
Also Read:
- Anthropic Accelerates Global Expansion as International Demand for AI Models Surges
- United Nations Inaugurates Global Dialogue on AI Governance to Shape Future of Artificial Intelligence
Legal experts note that while the $3,000 per-book figure provides immediate certainty, it is significantly higher than typical statutory damages, being four times larger than the $750 floor. This resolution is expected to have far-reaching implications, influencing how other major AI developers, including Meta and OpenAI, navigate their own ongoing copyright infringement lawsuits and acquire training data moving forward.
