NEW YORK / RankWire.AI / – Hachette Book Group, Cengage Learning and Elsevier sued Google over its Gemini artificial intelligence models. Author Scott Turow and his company, S.C.R.I.B.E., joined the proposed class action filed July 10 in federal court in Manhattan. The plaintiffs accuse Google of copying millions of books and journal articles without permission. They say the company used those works to develop and train Gemini. The complaint seeks damages, an injunction and a jury trial.
The 57-page filing presents four claims under federal copyright law. Three claims address reproduction through Google services, large-scale web scraping and model training. The fourth invokes the Digital Millennium Copyright Act. The plaintiffs allege Google removed or changed copyright management information from works included in training data. For this case, Gemini includes versions and relatives of LaMDA, PaLM, Bard and the current Gemini models. The court has made no liability finding and has not certified the proposed class.
The complaint focuses partly on Google Books, Google Play Books and Google Scholar. Publishers and authors supplied works for limited search, sales or research functions, according to the filing. The plaintiffs say those arrangements did not authorize commercial AI training. They also allege Google gathered copyrighted material through broad web scrapes, including content from pirate sites and paywalled services. The complaint cites an internal assessment that estimated potential fines between $10 billion and $100 billion. The court has not tested that evidence.
Proposed class targets registered works
The proposed class covers owners of registered U.S. copyrights in qualifying books and journal articles. Covered books must carry an International Standard Book Number, commonly called an ISBN. Covered articles must have a Digital Object Identifier or an International Standard Serial Number. The definition includes works allegedly copied from Google services, scraped online or reproduced during Gemini development and training. Registration dates and exclusions in the complaint would determine which copyright owners qualify. The court must approve class status.
The filing lists sample works controlled by each plaintiff. Hachette identifies titles including Scott Turow’s novel Innocent and a children’s book by Lemony Snicket. Cengage points to textbooks covering psychology, economics and other fields. Elsevier identifies scholarly papers published in medical and scientific journals. The plaintiffs say Google copied representative works through multiple routes and repeated that copying during model training. They seek discovery to identify the full collection of affected material. The complaint does not state a total damages figure.
Publishers seek accounting and damages
The requested remedies include statutory damages or actual damages and profits tied to any infringement proven in court. The plaintiffs also seek an order stopping unlawful copying. They ask Google to disclose Gemini training materials, collection methods and known model capabilities. The proposed accounting would identify copyrighted works and explain how Google collected, processed and stored them. The complaint also requests court-supervised destruction of unauthorized copies under Google’s control. It seeks legal costs and other relief allowed by law.
The case carries the number 1:26-cv-05870 in the Southern District of New York. It follows an earlier attempt by Hachette and Cengage to join separate Google AI copyright litigation in California. The publishers say the New York action preserves claims outside the proposed class in that proceeding. The new complaint adds Elsevier, Turow and S.C.R.I.B.E. alongside the two earlier participants. It asks the court to assess Google’s conduct under the Copyright Act and the Digital Millennium Copyright Act. Google remains the sole defendant in the New York case.