Major Publishers Sue Google Over Gemini Training — 'One of the Most Prolific Infringements in History'

Three major publishers and bestselling author Scott Turow have filed a federal lawsuit against Google, alleging the company illegally used millions of copyrighted books to train its Gemini AI models in what the complaint calls “one of the most prolific infringements of copyrighted materials in history.”

🔍 THE BOTTOM LINE

Google gave publishers a deal: let us scan your books for Google Play Books and Google Scholar, and we’ll display snippets and sell ebooks. The publishers say Google then took those same scanned books and fed them into Gemini’s training pipeline — a use no one agreed to. According to the complaint, Google’s own internal documents acknowledged the risk: “$10Bs-$100Bs in potential fines.” They did it anyway. This case isn’t about whether AI training is fair use. It’s about whether a company can repurpose content it obtained under one licence for an entirely different, unlicensed purpose.

The Allegations

The lawsuit, filed in federal court in New York, was brought by Hachette Book Group, Cengage Learning, Elsevier, and Scott Turow, the bestselling author of Presumed Innocent.

According to the Guardian’s reporting, the publishers argue that Google repurposed books originally supplied for limited services — Google Books, Google Play Books, and Google Scholar — where the licensing allowed specific uses like displaying searchable snippets or selling ebooks. Those licences did not, the lawsuit claims, extend to copying the works for training commercial AI products.

“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,” the suit states.

Google Knew the Risks

The most damaging detail in the complaint isn’t the infringement itself — it’s that Google allegedly knew the infringement was illegal and proceeded anyway.

According to the filing, internal Google discussions acknowledged the company could face “$10Bs-$100Bs in potential fines” for using texts provided by publishers for Google Play Books in AI training. The complaint claims Google made copies of copyrighted books to train Gemini without permission or payment, despite these internal warnings.

The publishers argue that AI-generated content could negatively impact book sales. The complaint offers a concrete example: Gemini could generate “a 100-page murder mystery set in a quiet seaside town filled with secrets, that substitutes for an original copyrighted murder mystery on which Gemini trained” in 20 minutes for 39 cents. “No publisher or author can compete with that.”

Specific books named in the lawsuit as allegedly used without permission include NK Jemisin’s The Fifth Season and Lemony Snicket’s Who Could That Be at This Hour?

A Growing Legal Battlefield

This case is the latest in a series of copyright lawsuits targeting AI companies. The landscape includes:

A judge ruling in Meta’s favour in a copyright case brought by authors last June — a decision that publishers are now trying to distinguish from their own claims
Anthropic’s landmark $1.5 billion settlement with authors who alleged pirated copies of their books were used to train Claude
Our previous coverage of publishers suing Meta over Llama training, which raised similar legal questions about AI training data
Thousands of authors including Kazuo Ishiguro and Richard Osman publishing an “empty book” in protest earlier this year

What sets this case apart from the Meta and Anthropic lawsuits is the breach-of-trust angle. The publishers didn’t just find their books scraped from the internet — they had a direct licensing relationship with Google for Google Play Books. Google had the books legitimately. The allegation is that it used them for something the licence never contemplated.

The Publishers’ Demands

The plaintiffs are seeking statutory damages, a permanent injunction preventing Google from continuing the alleged infringement, and a court order requiring the company to destroy any unauthorised copies of their works used in training its AI systems.

This is a significant ask. If granted, a destruction order would mean Google would have to either retrain Gemini from scratch without the publishers’ content or demonstrate that the works could be surgically removed from the model — something no AI company has yet proven is technically possible.

Google did not respond to a Guardian request for comment.

Why This Case Could Set the Precedent

The Meta copyright case last June hinged on whether scraping publicly available books constituted fair use. The court said yes. But this case asks a different question: when a company obtains copyrighted material under a specific licence for one purpose, can it use that material for an entirely different purpose without a new licence?

If the court rules for the publishers, it establishes that AI training is a distinct use requiring separate licensing — even when the underlying content was already legally obtained. That would upend the business model of every AI company that has repurposed its existing content libraries for training.

The Australian AI copyright landscape has been grappling with the same questions, with the Albanese government attempting to balance AI innovation against creator rights. NZ’s Copyright Act 1994 has yet to be tested on AI training data, but the Ministry of Business, Innovation and Employment has been monitoring international cases closely.

❓ FAQ

How is this different from the Meta copyright lawsuit? The Meta case involved books scraped from the internet. This case involves books that Google obtained through a direct licensing relationship with publishers for Google Play Books and Google Scholar. The publishers allege Google repurposed that licensed content for AI training — a use the licence didn’t cover. The legal question is about licence scope, not fair use of public material.

What are the publishers asking for? Statutory damages, a permanent injunction against further infringement, and a court order requiring Google to destroy unauthorised copies used in training. A destruction order would be technically unprecedented — no AI company has demonstrated it can remove specific training data from a trained model.

Did Google know this was legally risky? According to the complaint, yes. Internal Google discussions allegedly acknowledged “$10Bs-$100Bs in potential fines” for using Google Play Books content in AI training. The publishers will argue this shows wilful infringement, not a good-faith interpretation of fair use.

What books are named in the lawsuit? The complaint specifically mentions NK Jemisin’s The Fifth Season and Lemony Snicket’s Who Could That Be at This Hour? as examples of copyrighted works allegedly used without permission for Gemini training.

How does this affect New Zealand publishers and authors? NZ publishers who licensed content to Google Play Books could potentially join or file similar claims. The Copyright Act 1994 doesn’t yet have AI-specific provisions, but existing copyright protections would apply. NZ authors whose works appeared on Google Play Books should check whether their licensing agreements included AI training as a permitted use.

🔍 THE BOTTOM LINE

Google had the books legitimately. It had licensing agreements with the publishers for Google Play Books. What it allegedly didn’t have was permission to use those books to train Gemini — and its own internal documents reportedly acknowledged the exposure was in the tens of billions. The precedent here matters enormously: if repurposing licensed content for AI training is a separate infringing use, every AI company that has mined its own content libraries is sitting on the same liability. The Anthropic settlement showed that paying up is one way out. Google, so far, is choosing to fight.