The US publishers Hachette, Elsevier, and Cengage, as well as bestselling author Scott Turow, filed a class action lawsuit against Google on July 13, 2026, in a federal court in New York. The company allegedly used millions of copyrighted books and scholarly articles without permission to train its AI model Gemini, partly from illegal sources on the internet.
The complaint lists specific allegations against Google
The complaint, drafted by the law firms Oppenheim + Zebrak and Keller Rohrback, accuses Google of misusing books from the Google Books program: according to the publishing group Hachette Book Group, these were originally released only for limited purposes such as full-text search. Additionally, the company allegedly scraped large parts of the internet, including content from well-known piracy sources and from paid offerings behind paywalls.
According to the lawsuit, Google also removed so-called copyright management information from the works to obscure the origin of the training data. Gemini also generates content that could directly replace original works: the model reportedly produced a summary of about 2,000 words of Turow’s novel “Innocent,” according to TechCrunch. Such output stands in direct competition with official study guides and book reviews, the plaintiffs argue.
Textbook publisher Cengage also reports cases in which Gemini reproduced entire chapters of academic works nearly verbatim. The plaintiffs interpret this as evidence that Google far exceeded the usage rights originally granted.
Internal document warns of billion-dollar fines
A central piece of evidence in the lawsuit is an internal Google document that classified the use of copyrighted books for AI training as “highly problematic.” The analysis reportedly warned of potential fines ranging from ten to one hundred billion dollars – a figure that has not been independently verified and comes solely from the complaint itself.
This case is not the first of its kind: the same four plaintiffs already filed a nearly identical lawsuit against Meta in May 2026. Two federal judges in California had previously ruled in separate proceedings that training AI models with copyrighted material can generally fall under the fair use doctrine. However, the plaintiffs in the current case argue that Google also violated contractual terms and the prohibition on removing copyright information, which goes beyond a pure fair-use question.
How costly an AI copyright dispute can become for a company was demonstrated recently by Anthropic, Adweek reports: the company paid 1.5 billion dollars to settle a similar class action lawsuit – so far the highest known settlement amount in a US copyright case. Google has not publicly commented on the new allegations.
What will be decisive is whether the court classifies the training practice as fair use despite the additional allegations surrounding removed copyright information and alleged breach of contract, as American federal judges have already done in the cases against Meta and OpenAI. The court has not yet scheduled a date for an initial hearing; given comparable proceedings, a decision could take several months.
