Several of the world's largest book publishers have taken Google to court, accusing the company of using millions of copyrighted books without authorization to train its Gemini AI models.
The publishers Hachette Book Group, Cengage and Elsevier, joined by the novelist Scott Turow, filed a proposed class-action lawsuit in the US District Court for the Southern District of New York, according to Publishers Weekly. The suit seeks damages and a court order to stop what the plaintiffs describe as large-scale copyright infringement.
The allegations
The complaint alleges that Google drew on copyrighted works from a range of sources to train Gemini, including books made available under earlier, limited arrangements for services such as Google Books, as well as material the publishers say was taken from pirated sources, as TechCrunch reported. The plaintiffs contend that Google repurposed content it was allowed to use only for narrow purposes, such as displaying short snippets, and used it instead to build commercial AI systems.
The Association of American Publishers, which is backing the case, said the alleged copying was "willful" and threatened the market for books by enabling cheaper, machine-generated substitutes and undercutting emerging licensing deals for training data.
Google's position
Google has not issued a detailed public response to the suit. The company has argued in other cases that training AI systems on large bodies of text can be a "transformative" use protected under US copyright law, a defense the industry has leaned on heavily as such disputes multiply.
Part of a wider fight
The lawsuit is one of many now testing how copyright law applies to generative AI. Authors, news organizations, music companies and visual artists have brought a wave of cases against AI developers including OpenAI, Meta and Anthropic. In a closely watched settlement in 2025, Anthropic agreed to pay $1.5 billion to resolve claims brought by authors, NPR reported, in a case where a court had found that training on lawfully acquired books could be fair use, but that storing pirated copies was not.
Courts have so far reached mixed conclusions, often distinguishing between how AI models learn from text and how the underlying material was obtained. The outcome of the publishers' case against Google could help shape whether, and how, companies must pay for the vast troves of copyrighted writing that power modern AI, an unsettled question with billions of dollars and the future economics of publishing at stake.
