Publishers sue Google over Gemini AI training on copyrighted books

Publishers handed Google their catalogs years ago just to show search snippets and sell physical books. Now, they are accusing the search giant of turning those very pages into high-octane fuel for its flagship AI.

A heavyweight coalition of publishers including Hachette, Cengage, and Elsevier, alongside author Scott Turow and writers group S.C.R.I.B.E., has filed a major class action lawsuit against Google in the U.S. District Court for the Southern District of New York. The claim is simple yet damning: Google used copyrighted books to train Gemini without permission, and allegedly scrubbed copyright metadata to cover its tracks.

The $100B internal warning

This dispute cuts deeper than a standard web scraping fight because the publishers had an established, scope-limited relationship with Google.

For decades, authors provided access so their titles could appear in Google Books, where readers were strictly limited to brief bibliographic snippets. The lawsuit alleges Google ignored those contractual guardrails, lifting protected books from both Google Books and the Google Play store to feed Gemini.

Even worse for Google's defense, the lawsuit cites an internal document where company insiders allegedly warned that training on copyrighted books was highly problematic and could trigger potential fines between $10 billion and $100 billion.

Moving past the California playbook

The legal showdown comes at a delicate time for copyright law in generative AI.

Two early rulings out of California leaned in favor of Big Tech, treating AI training as fair use under statutes written long before modern machine learning. But Anthropic already took a massive hit, paying out a record $1.5 billion settlement over pirated training data where hundreds of thousands of writers qualified for checks.

Many authors explicitly opted out of that Anthropic payout to keep their litigation options open. By filing in Manhattan, the publishers are deliberately testing a new bench that may view Google's conduct through a much harsher commercial lens.

Why it matters

If the allegations in this lawsuit hold up, Google's legal exposure is terrifying.

We think Google made an unforced strategic blunder here. Instead of utilizing its massive balance sheet to strike landmark licensing deals and lock up definitive, high-grade training corpuses away from competitors, it seemingly cut corners with materials it already stored for other purposes.

For content creators, this lawsuit is about establishing an undeniable financial toll for bypassing licensing. Even if individual royalty checks end up modest, a win here will dismantle the free-for-all web harvesting model.

For the broader AI ecosystem, the era of assuming pre-training is automatically fair use looks like it is running out of road. If federal courts in New York split from California, AI developers may soon face mandatory, industry-wide data clearinghouses that significantly raise the barrier to entry.