Anthropic settlement raises new pressure over AI training data
The Anthropic settlement in a major authors’ copyright lawsuit has received final approval from a U.S. federal judge, closing one of the most closely watched AI training data cases so far. The Hindu reported the approval, while Reuters said U.S. District Judge Araceli Martinez-Olguin signed off on the $1.5 billion deal on July 20, 2026. Reuters described it as the largest known settlement in a U.S. copyright case.
The case, known as Bartz v. Anthropic, was brought by authors who accused Anthropic of using pirated books to train Claude, its AI chatbot. According to AP, the lead plaintiffs included thriller writer Andrea Bartz and nonfiction authors Charles Graeber and Kirk Wallace Johnson. The lawsuit later expanded to cover a larger class of authors and publishers.
This Anthropic settlement is not a simple ruling against AI training. In June 2025, Judge William Alsup ruled that training AI on books could qualify as fair use if the books were obtained legally. But he also found that Anthropic could face liability for storing millions of pirated books in what Reuters described as a central library. That split is why the Anthropic settlement matters: the court treated training and piracy as separate issues.
The official Anthropic Copyright Settlement website says class members include legal or beneficial copyright owners of books covered by the settlement. It also says the claims deadline passed on March 30, 2026, while opt-out and objection deadlines passed on February 9, 2026.
The money is large, but the structure is specific. The Authors Guild said the $1.5 billion fund would be divided among rightsholders for eligible books after fees and expenses. It said about 500,000 titles were expected to qualify, implying at least $3,000 per title before deductions.
Anthropic framed the deal as a way to close the remaining claims. Reuters quoted Anthropic deputy general counsel Aparna Sridhar as saying the company reached the settlement after the court’s ruling that AI training on books is fair use under copyright law. Plaintiffs’ lawyer Justin Nelson called it a historic settlement, according to Reuters.
The approval could shape lawsuits against other AI companies, including cases involving OpenAI, Microsoft, Meta and Midjourney. The central question is whether courts will keep drawing a line between legal training use and unlawful data collection. For related coverage on AI companies and legal pressure, read The AI Decode’s reporting on Anthropic valuation and AI regulation.
For authors, the Anthropic settlement offers payment after a long dispute. For AI companies, it sends a narrower warning. Training may survive fair use arguments in some cases, but the source of training data is becoming harder to ignore.
