Publishers and Authors Get First Major AI Copyright Settlement

This month marked a major decision in the ongoing fight between publishing companies and artificial intelligence. A class action lawsuit against AI developer Anthropic reached a settlement of $1.5 billion. Under the settlement, members of the class action will get $3,000 per written work included in the lawsuit, split equally between publisher and author. 

British publisher Bloomsbury Publishing confirmed last week that it was a member of that class action. Bloomsbury had 14,087 books involved in the case, for a total payout of about $42 million—after splitting with authors, the company stands to get $21 million. This is considered to be the biggest copyright payout in U.S. legal history. 

The lawsuit first started in August of 2024, when several authors—Andrea Bartz, Kirk Wallace Johnson, and Charles Graeber—sued the Claude chatbot developer for copyright infringement. 

“Anthropic has built a multibillion-dollar business by stealing hundreds of thousands

of copyrighted books,” that first complaint stated. “Rather than obtaining permission and paying a fair price for the creations it exploits, Anthropic pirated them.…An essential component of Anthropic’s business model—and its flagship ‘Claude’ family of large language models (or “LLMs”)—is the large-scale theft of copyrighted works.”

This kind of argument has been made by creatives across all fields since the advent of widespread generative AI, with concerns about its alleged pervasive, illegal scraping of human-produced creative works to train new large language models, without either obtaining permission from or fairly compensating the people who produced that training material. 

Anthropic’s position is that training AI on published books is “fair use,” according to a statement provided to NPR. It said it was “looking forward to bringing this matter to a close.”

This is not the only lawsuit of its kind. A number of others are still in the pipeline. A group of book publishers including global publishing giant Hachette Book Group sued Google's Gemini AI just two weeks ago for copyright infringement; in May, five major publishers sued Meta for the same reason, alleging it had illegally scraped their published books to train its Llama AI models. 

Publishers have also taken aim at the indirect feeders of AI copyright infringement, with major lawsuits against open-access “shadow libraries” like Anna’s Archive, a May lawsuit which was quickly and favorably ended, and WeLib, a lawsuit that was started just one month later. These shadow libraries offer millions of copies of books, academic articles, and other documents, and often partner with AI companies to provide them as training data for a fee. 

In the approval of the final Anthropic class action settlement, released on July 20, U.S. district court judge Araceli Martínez-Olguín wrote that the money would bring “meaningful relief” and “substantial benefits” to the plaintiffs, and that the payout was “fair, reasonable, and adequate.” So far, about 91 percent of class members have claimed their share of the payout; Bloomsbury told Reuters that payments would begin in the second half of the financial year.