Publishers sue Google over Gemini AI training, calling it one of history's biggest copyright violations

Key points

• Hachette Book Group, Cengage Learning, and Elsevier filed a federal lawsuit against Google in New York.
• Bestselling author Scott Turow joined the three publishers as a plaintiff.
• The suit accuses Google of using millions of copyrighted books without permission to train its Gemini AI models.
• The plaintiffs describe it as "one of the most prolific infringements of copyrighted materials in history."
• The case joins a wave of copyright litigation targeting AI developers over their training data.

Three major book publishers and celebrated American novelist Scott Turow are taking Google to federal court, accusing the tech giant of copying millions of books to teach its artificial intelligence how to write and reason.

The case, first reported by The Guardian, was filed in federal court in New York. It targets Gemini, Google's family of large language models (the technology that powers AI assistants). Training on books is especially valuable because they contain long, carefully constructed prose, qualities that help an AI respond coherently. None of that copying was authorised, the publishers say.

AI2Day has tracked the broader copyright fight since the site's earliest days, including Australia's creatives pushing for legislative protection on 13 July 2026. This New York filing shows the battle has moved firmly into US federal courts.

"One of the most prolific infringements of copyrighted materials in history" is the phrase used in the complaint. OpenAI, Meta, and similar companies face comparable suits from authors and news organisations who say their work was taken without consent or payment. But this coalition is notable: Hachette covers popular fiction and non-fiction, Cengage produces educational textbooks, and Elsevier publishes scientific and medical journals. That breadth suggests something beyond a niche grievance.

What does this mean for people who read or buy books?

For now, nothing changes on your shelf or your e-reader. This is a civil dispute about how Google built its technology, not a product recall.

The outcome could matter quite a lot down the line, though. If courts decide AI companies must licence the books they train on, publishers and authors gain a new revenue stream. Prices for AI products could rise to cover those costs, or companies might train on smaller, licensed datasets, which could affect how capable the next generation of AI tools turns out to be.

Google hasn't filed a public response to the complaint. The company has previously argued, in other copyright cases, that training AI on publicly available text falls under "fair use," a legal doctrine allowing limited use of copyrighted material without permission under defined conditions. Federal courts haven't settled the question.

Should you worry about AI copyright cases piling up?

These cases are worth watching precisely because there are so many of them and none has produced a definitive ruling. What's being somewhat oversold here is urgency: the legal process will take years. What's genuinely at stake is whether the economics of AI training get renegotiated from scratch, and whether the answer comes from courts or from Congress.