Publishers sue Google, cite internal memo warning of $100B in AI fines

A heavyweight coalition of publishers and authors—including Hachette, Cengage, Elsevier, and novelist Scott Turow—has dragged Google into federal court, accusing the company of feeding their copyrighted books to its Gemini AI models without permission. The class action, filed in the Southern District of New York, alleges Google not only ignored licensing boundaries but actively stripped copyright management information to hide what the suit calls training on “stolen materials.”

This isn’t the usual scattershot copyright gripe. The plaintiffs argue they had a carefully defined relationship with Google through programs like Google Books and the Play store—deals that let the company display snippets and sell copies, not ingest entire texts into training pipelines. The complaint quotes an internal Google document that reportedly flagged the practice as “highly problematic” and warned of potential fines ranging from $10 billion to over $100 billion. That’s a staggering number, even by Big Tech standards, and it suggests someone inside the company saw the legal cliff before they drove off it.

The venue matters here. Two California federal judges have already sided with AI companies on fair use grounds, leaning on a copyright statute that predates the commercial internet. But this lawsuit lands in New York, giving a different judge a crack at the question. The publishers are betting that Google’s pre-existing contracts and the alleged destruction of copyright metadata create a fact pattern those California rulings didn’t fully contemplate. It’s a smart play—arguing breach of a specific business relationship hits different than a blanket claim that all training data requires a license.

Anthropic’s recent $1.5 billion piracy settlement looms in the background, though half a million writers qualified for payouts and many walked away to keep suing. The appetite for a definitive ruling hasn’t dimmed. With Google’s own internal math suggesting the stakes could reach ten figures, the company’s silence so far is deafening. I’ll be watching whether discovery forces those documents into the open—because if that $100 billion figure proves authentic, the fair use shield might not be as sturdy as Silicon Valley hopes.