A new Google Gemini lawsuit is forcing a major question into the open: Can a technology company use copyrighted material to build artificial intelligence simply because it already has access to the files?
According to the class action complaint filed in the Southern District of New York, Hachette Book Group, Cengage Learning, Elsevier, author Scott Turow and his company S.C.R.I.B.E. accuse Google of copying protected books and scholarly articles without permission to develop and train Gemini. The lawsuit was filed on July 10, 2026, and seeks to represent a larger group of authors and publishers whose work was allegedly used.
For anyone unfamiliar with AI training, the process starts with massive collections of information called training data. A company feeds books, articles, websites and other material into a large language model so the system can study patterns in language. The model learns how words, ideas and structures relate, allowing it to answer questions or generate new text when a user enters a prompt. According to Google’s Gemini technical report, the technology was designed to process and reason across text, images, audio and video.
That process requires an enormous amount of quality material. Random internet posts can help a model learn basic language, but professionally edited books, textbooks and research papers offer organized information, strong writing and specialized knowledge. That makes publisher libraries extremely valuable to companies competing to build smarter systems.
The lawsuit is different from a basic complaint about material found on the public internet. According to the court filing, publishers had already given Google access to clean digital copies of their work through programs created for specific purposes. Google Books made books searchable, Google Play Books sold authorized electronic editions and Google Scholar helped users locate academic research. The plaintiffs say none of those arrangements authorized Google to reuse the files for commercial AI training.
The publishers involved cover multiple parts of the industry. According to Hachette Book Group, it publishes general interest titles across fiction, nonfiction, children’s literature and several other categories. Cengage says it provides educational materials and digital learning products for students and professionals. Elsevier describes its business as supplying scientific and healthcare information, including research journals. Together, the companies control the rights to material ranging from novels to textbooks and technical studies
According to the complaint, Google internally described using publisher supplied books for AI as “highly problematic for Google” and warned of “$10Bs-$100Bs in potential fines.” The filing also attributes a blunt explanation to Gemini’s lead engineer: “we don’t do deals for data we already have or already possess.” Those statements remain allegations presented by the plaintiffs, not findings that a judge or jury has confirmed.
The filing also reaches beyond files supplied directly by publishers. According to the complaint, Google allegedly built training collections using large internet scrapes that included material from pirate sites and content taken from behind subscription barriers. The plaintiffs further claim that Google removed or altered copyright management information, including author names, ownership details and publication data, making the original works harder to trace. Google has not been found liable for those allegations.
Why does this matter to working writers? The publishers argue that an AI system can absorb human created books and then produce cheap alternatives at a speed no author or editorial team can match. According to the complaint, Gemini was prompted to create a 100 page murder mystery in about 20 minutes at a reported cost of 39 cents. The filing presents that example as evidence that AI products could flood the market with books, replacement textbook chapters and imitations competing against the same creators whose work allegedly helped train the system.
That concern carries extra weight for Black writers and independent creators. Publishing has never offered equal access, marketing support or financial protection to every voice. A system that can study a creator’s language, cultural references and storytelling choices without a clear licensing process could widen an existing power gap. The issue is not whether creators should fear technology. It is whether the companies earning money from that technology must negotiate with the people whose labor helped make it useful.
The case does not automatically mean that training an AI model with copyrighted work is illegal. Courts are still sorting out when AI training may qualify as fair use, a doctrine that can permit certain unlicensed uses of protected material. According to the Authors Guild’s summary of the Anthropic litigation, a federal judge found that training on legally acquired books could qualify as fair use while allowing claims tied to pirated copies to continue. Anthropic later agreed to a $1.5 billion settlement covering hundreds of thousands of books, according to the Associated Press.
That distinction could become central to the lawsuit. The publishers are not only challenging what Gemini learned. They are challenging how Google allegedly obtained and reused the material. They argue that permission granted for search, discovery or retail sales cannot silently become permission to build a separate product worth billions.
According to the complaint, the plaintiffs want class certification, monetary damages, an order stopping unlawful copying, an accounting of Gemini’s training materials and the destruction of unauthorized copies. The Guardian reported that Google did not respond to its request for comment following the filing. Until Google answers the allegations in court, the publishers’ claims remain one side of a legal fight that could help decide who gets paid when human creativity becomes raw material for artificial intelligence.
