If you think the legal battles over AI training are just about web scrapers sucking up public blog posts, you're missing the real story.
A massive class-action lawsuit filed in New York federal court on July 10, 2026, has shifted the entire copyright debate. Three publishing giants—Hachette Book Group, Cengage Learning, and Elsevier—along with best-selling novelist Scott Turow, are taking Google to task.
They aren't just accusing Google of mindless web scraping. They're accusing the tech giant of breaking its own platform agreements by taking digital books from its Google Books, Google Play, and Google Scholar services, and feeding them directly into Gemini.
Even worse for Google, the lawsuit highlights internal warnings that show the company knew exactly how dangerous this move was.
The smoking gun in the publisher lawsuit
The most damaging detail in the legal filing is a private internal memo. Long before this lawsuit hit the docket of the Southern District of New York, Google’s own staff flagged the risks of using publisher-provided digital books for training artificial intelligence.
The internal analysis warned that training on these books was "highly problematic". It even predicted that publishers would see this as copyright infringement and could file lawsuits, risking "$10Bs-$100Bs in potential fines".
They did it anyway.
The plaintiffs allege that Google deliberately took books and academic articles from restricted, scope-limited programs. According to the complaint, Google then stripped copyright management data from these files to hide the evidence that copyrighted material was being used to build Gemini’s core capabilities.
This isn't just a simple mistake. The publishers are calling it willful infringement on a massive scale.
Why Google Books was never meant for this
To understand why publishers are so angry, we have to look back at the history of Google Books.
Decades ago, Google scanned millions of physical books to build a searchable index. Publishers fought back then, too. The court eventually ruled in Google’s favor, but only because of a very narrow, highly specific compromise: Google could display tiny snippets of text so users could find books, which would ultimately drive sales. It was deemed "fair use" because it didn't replace the book.
This lawsuit argues that Google has completely smashed that compromise.
Using those same digital copies to train an AI that writes its own stories is a completely different ballgame. The books aren't being indexed for search anymore. They are being used as training material to build a tool that aims to replace human authors.
The lawsuit states that Gemini can generate a 100-page murder mystery set in a quiet seaside town in 20 minutes.
The cost? A measly 39 cents.
No human author can compete with those economics. When a system can churn out near-verbatim book chapters, textbook replacements, and detailed novel summaries, it ceases to be a search engine. It becomes a competitor.
The shift to New York is bad news for tech giants
For the last couple of years, tech companies have enjoyed a bit of a home-court advantage. Several early AI copyright rulings in California federal courts sided with technology firms, finding that training on copyrighted data fell under the umbrella of "fair use".
But the publishers and Scott Turow have pulled a smart geographic move. They officially withdrew their bid to join an ongoing California class action and filed this fresh lawsuit in New York’s Southern District instead.
The New York court is not bound by those California rulings.
The legal wind is also starting to blow in a different direction. Recently, Anthropic was hit with a staggering $1.5 billion penalty for training its models on copyrighted works, which allowed half a million authors to claim minimum statutory damages. That case proved that courts are losing their patience with the "move fast and break things" approach to intellectual property.
By centering the lawsuit in New York, the publishers are banking on a judicial circuit that is traditionally much more protective of intellectual property, copyright holders, and the creative industries.
How Google allegedly gathered the training data
The lawsuit divides Google’s training sources into three distinct, unauthorized buckets:
- Google Play and Partner Books: Digital files uploaded by publishers for the sole purpose of selling retail e-books to consumers.
- Google Books and Google Scholar: Databases built under strict, legally binding terms that limited their use to search indexing and academic reference.
- Pirate Sites and Paywalled Scrapes: Scraping the broader web, including known book piracy repositories and content hidden behind subscription paywalls.
The publishers aren't just suing for a payout. They want the court to declare Google's actions unlawful, block future infringement, award maximum damages, and order the destruction of all infringing copies in Google’s possession. That last part—destroying the copies—is a nightmare scenario for any AI company, as it could require them to delete or completely retrain their core models.
What you should do to protect your digital content
If you write, publish, or manage digital content, this legal battle proves you can't rely on tech platforms to respect the original intent of your licensing agreements. You need to take active steps to safeguard your work.
Update your terms of service immediately
If you sell digital books, PDFs, or educational materials on your own website, make sure your terms of service explicitly prohibit the use of your content for machine learning and AI training. Use clear, direct language.
Implement technical blocks
Don't make it easy for scrapers to access your work. Keep your most valuable written content behind a secure paywall or login screen. While the lawsuit notes Google bypassed some of these blocks, it still provides you with stronger legal footing if they have to actively break through a barrier to get your data.
Use standard robots.txt exclusions
Add specific user-agent blocks to your site's robots.txt file to keep major AI crawlers out. Block Google-Extended and other known AI bots to prevent them from scanning your web-based writings for future model updates.
Keep an eye on platform updates
Review the terms of use for any digital distribution platforms you use. If you sell books through online stores, read the fine print of your distribution contracts to see if they grant the platform the right to use your files for "internal product development" or "algorithmic improvement." Opt out of those clauses whenever possible.
