Google sued by major publishers over Gemini AI training

Hachette, Cengage, Elsevier, author Scott Turow, and S.C.R.I.B.E. have brought a class action lawsuit against Google in the U.S. District Court for the Southern District of New York, asserting that the company illegally used their copyrighted works to train Gemini. The plaintiffs contend that Google not only trained the AI model on copyrighted books without authorization but also deliberately removed or altered copyright information to obscure the fact that Gemini was built on what they describe as "stolen materials."

The core complaint centers on a breach of trust. Publishers and authors had established a long-standing relationship with Google through Google Books, a service that allows users to search for books and view limited snippets alongside bibliographic data—but not full text. These publishers provided Google access to their copyrighted works specifically for this searchable index. The lawsuit asserts that Google copied those same books, along with titles uploaded to Google Play, and fed them into Gemini training without ever requesting or obtaining permission to do so. "Google illegally copied works from all these scope-limited programs for AI training, knowing it lacked authorization to do so," the complaint states.

The case arrives in a crowded legal landscape. Publishers, authors, and other copyright holders have filed multiple lawsuits against AI companies including Meta, OpenAI, and Anthropic. Many of these cases remain pending. So far, two early decisions from California courts have ruled in favor of the AI companies, finding that using copyrighted material for AI training constitutes "fair use" under U.S. copyright law—a statute that predates the internet. However, Anthropic received a different outcome: the company was fined $1.5 billion（約2400億円） for copyright infringement related to its training data, marking the largest payout in U.S. copyright law history. Roughly half a million writers were eligible for payments of at least $3,000, but many authors chose to decline the settlement in order to pursue additional legal claims.

The lawsuit also highlights an apparent internal assessment of risk at Google. The plaintiffs cite an internal document from the company stating that using copyrighted books for AI training could be "highly problematic for Google" and might generate "$10Bs–$100Bs in potential fines." This memo undercuts any argument that the company lacked awareness of legal exposure. Google has not yet responded to requests for comment.

This lawsuit arrives amid a widening legal battle between copyright holders and AI companies. Google, Meta, OpenAI, and Anthropic have all faced complaints from publishers, authors, and other copyright holders over the use of protected works in AI training. The regulatory and legal landscape is still unsettled: two early California court rulings have sided with the AI companies, treating the training use as "fair use" under U.S. copyright law. However, a different outcome came for Anthropic, which was fined $1.5 billion（約2400億円） — the largest payout in U.S. copyright law history — though many authors chose to reject that settlement to pursue independent legal claims.

What distinguishes Google's case is the context of prior licensing. Publishers provided Google with copyrighted works under a specific, limited agreement: to make books searchable through Google Books, which restricts users to snippets and bibliographic information, not full texts. The lawsuit contends that Google repurposed these same works for a wholly different purpose — AI model training — without seeking fresh permission. The company's own alleged internal assessment, cited in the complaint, suggested this could expose Google to massive financial liability. The filing in New York's federal court, rather than California, means a fresh judicial interpretation of fair use doctrine is possible, making the outcome less predictable than the California precedents might suggest.