Publishers and Authors File Class-Action Lawsuit Against Google Over Gemini AI Training

The tension between Silicon Valley's insatiable appetite for data and the publishing industry's right to control its intellectual property has finally spilled over into a New York federal court. For years, tech giants have scraped the internet to train the artificial intelligence models that now write code, draft emails, and generate novels on command. But a coalition of major book publishers and authors is drawing a hard line, arguing that the era of "move fast and ingest everything" must end. On July 10, 2026, Hachette Book Group, Cengage Learning, Elsevier, and bestselling author Scott Turow filed a class-action lawsuit against Google, accusing the company of executing "one of the most prolific infringements of copyrighted materials in history" to build its Gemini AI models.[1][4]

The lawsuit strikes at the core of how modern generative AI is developed. According to the complaint, Google did not simply scrape the open web; it allegedly repurposed millions of books and journal articles that publishers had provided for strictly limited services, such as Google Books and Google Play Books. Those agreements were designed to allow searchable snippets or digital retail distribution, not to serve as the foundational training data for a commercial AI product. By feeding these curated, professionally edited texts into Gemini, the plaintiffs argue, Google bypassed the very contracts it signed to gain access to the material in the first place.[2][3][5][7]

What makes this particular filing explosive is the inclusion of internal Google communications. The publishers' complaint cites documents suggesting that Google engineers were acutely aware of the legal risks involved in using publisher-provided texts. According to the filing, internal memos flagged the practice as "highly problematic," warning that secretly training on Google Play Books could expose the company to anywhere from $10 billion to $100 billion in potential fines. The documents allegedly show Google acknowledging that publishers were highly sensitive about their data and would likely view the unauthorized ingestion as blatant copyright infringement.[1][6][7]

Beyond the breach of contract claims, the lawsuit accuses Google of actively covering its tracks. The plaintiffs allege that Google violated the Digital Millennium Copyright Act by stripping or altering copyright management information from the digital files before feeding them into the Gemini training algorithms. This removal, the publishers argue, was a deliberate attempt to obscure the origins of the training data and hide the fact that copyrighted works were being used to teach the AI how to mimic human writing styles and synthesize complex information.[2][3][4]

The stakes for the publishing industry are existential. Generative AI models like Gemini are not just search engines; they are capable of producing long-form text that directly competes with the authors whose work trained them. The lawsuit notes that Gemini can generate a 100-page murder mystery set in a quiet seaside town in a matter of minutes, effectively substituting for the original copyrighted novels it ingested. For authors like Scott Turow, whose legal thrillers rely on intricate plotting and curated facts, the prospect of an AI instantly generating knockoff narratives represents a direct threat to their livelihoods.[1][4][5]

Google has historically defended its data collection practices under the doctrine of "fair use," arguing that training AI on copyrighted text is a transformative process that creates entirely new tools rather than derivative works. The company previously won a landmark copyright case regarding the digitization of books for its search engine, where a federal appellate court ruled that displaying snippets of text was legally permissible. However, the publishers contend that ingesting complete works to train a commercial generative AI system that competes with the original authors falls far outside the boundaries of that precedent.[4][6]

The choice of venue is also a strategic calculation. While recent California court rulings have broadly favored AI companies by treating training data ingestion as fair use, this lawsuit was filed in the U.S. District Court for the Southern District of New York. Legal observers note that the New York courts may offer a fresh judicial interpretation of how copyright law applies to generative AI, potentially setting up a circuit split that could eventually force the Supreme Court to weigh in on the issue.[2][4]

This lawsuit is part of a broader, industry-wide reckoning over AI training data. It follows a wave of similar litigation against companies like OpenAI, Meta, and Anthropic. Just weeks before this filing, a federal judge finalized a landmark $1.5 billion class-action settlement between Anthropic and hundreds of thousands of writers over the unlicensed use of digitized books to train its Claude chatbot. That settlement has emboldened the publishing industry, signaling that courts are increasingly willing to hold AI developers financially accountable for the data they consume.[1][2][5]

As the case moves forward, it will test the legal and economic foundations of the generative AI boom. If the publishers succeed in securing an injunction or massive statutory damages, it could force tech giants to fundamentally alter how they source their training data, shifting the industry toward a licensing model where creators are compensated for their contributions. Until then, the lawsuit stands as a stark declaration from the publishing world: the raw material of the AI revolution is not free for the taking.