Publishers File Copyright Class Action Against Google’s Gemini

Publishers target Gemini training data

Hachette Book Group, Cengage Learning, Elsevier, author Scott Turow and writers’ organisation S.C.R.I.B.E. have filed a proposed class action lawsuit against Google, alleging the company copied copyrighted books and other works to train its Gemini AI platform without permission or payment.

The complaint was filed on 10 July in the US District Court for the Southern District of New York and became publicly known this week. It brings three claims under the Copyright Act and a fourth under the Digital Millennium Copyright Act, or DMCA.

The claimants are seeking statutory damages, permanent injunctions to prevent further alleged infringement, disclosure of Gemini’s training sources and court-supervised destruction of copies they say Google obtained without authorisation. Google had not publicly responded to the new lawsuit when reports emerged.

At its centre is a dispute over what Google was allowed to do with books supplied for existing products, including Google Books, Google Play Books and Google Scholar. Those services allow searchable snippets or ebook sales, but the publishers argue that the arrangements did not permit Google to reuse the works for training a commercial generative AI system.

Complaint alleges use of pirate libraries and paywalled material

The filing alleges Google downloaded texts from known pirate sources including Z-Library, OceanofPDF and WeLib. It also claims the company scraped material from behind paywalls and subscription platforms.

The plaintiffs allege that Google “reproduced millions of copyrighted works without permission, without providing any compensation to authors or publishers, and with full knowledge that its conduct violated copyright law.” They describe the alleged activity as one of the most extensive copyright infringements in history.

The complaint cites internal Google material that, according to the plaintiffs, warned using copyrighted books for AI training could be “highly problematic for Google” and expose the company to “$10Bs–$100Bs in potential fines”. It also attributes a remark to Gemini’s lead engineer that Google does not make deals for data it already has or possesses.

These are allegations made in the lawsuit, not findings by a court. A central question will be whether Google had the necessary rights to use works acquired for its book-search and sales services in a separate AI training process.

Publishers say Gemini can replace paid access

The claimants argue that Gemini is not simply analysing books in the abstract, but can generate material that competes with the original works. The complaint says the system can produce a 100-page murder mystery novel in around 20 minutes for roughly $0.39.

It also cites tests in which Google’s AI generated a textbook table of contents, reproduced a passage and produced an approximately 2,000-word summary of Turow’s novel Innocent. The plaintiffs contend that responses at that level of detail could substitute for buying a book or paying for access to it.

Named works in the filing include NK Jemisin’s The Fifth Season and Lemony Snicket’s Who Could That Be at This Hour?. The case is therefore aimed not only at the original copying alleged in the complaint, but at the potential market effect of AI-generated outputs.

“Desperate to maintain its online dominance, Google abandoned its early motto of ‘Don’t be evil’ and engaged in one of the most prolific infringements of copyrighted materials in history.”

A new venue for the AI copyright fight

This is not the first major legal challenge over AI training data. Hachette and Cengage had previously sought to join the Google Generative AI Copyright Litigation, a case first brought by illustrators and writers in 2023. They have now withdrawn that motion, noting that Google could argue a three-year statute of limitations applies to claims by class members.

The New York case also follows a May lawsuit against Meta in the same Manhattan federal court. That action involved a near-identical group of claimants, including Elsevier, Cengage, Hachette and Turow, and alleged Meta pirated millions of works to train its Llama models.

Elsewhere, the legal picture remains unsettled. A group of authors lost a 2025 case against Meta after a federal judge ruled that the AI training at issue met the requirements for fair use. Authors pursuing Anthropic reached an initial $1.5 billion settlement in 2025, but the judge overseeing it rejected the proposed deal as incomplete.

Because this Google action was filed in New York rather than California, another federal court will now have an opportunity to consider how fair use applies to generative AI and whether Google’s existing book agreements covered the alleged activity.

Why UK publishing is watching

The dispute is American, but its effects could travel. The UK-based Society of Authors and the Publishers Association have described Google’s alleged conduct as unconscionable, while Publishers Association general counsel Catriona MacLeod Stevenson said more cases of this kind could follow in 2026.

Google has not struck licensing agreements with digital publishers, unlike Meta, Microsoft, Amazon, OpenAI and Anthropic, according to the background to the claim. If the publishers succeed, AI companies could face substantial licensing changes, data-destruction orders and a slower route to building future models.

For ordinary book buyers, the immediate question is whether cheap AI-generated substitutes will erode the market for authors and publishers whose work helped make those systems useful. UK publishers, writers and agents will be watching the New York proceedings closely for a precedent that could shape future European and British legal action.