Publishers challenge Google’s use of books to train Gemini
Major publishers and authors have sued Google over the alleged use of copyrighted books to train Gemini, raising questions about whether content licensed for one service can be reused for AI.
A new lawsuit against Google focuses on a particularly important AI copyright question: whether material provided for one limited purpose can later be reused to train a general-purpose model.
What happened
Publishers and authors including Hachette, Cengage, Elsevier and Scott Turow filed a proposed class action alleging that Google used copyrighted books to train Gemini without permission. The plaintiffs also allege that copyright-management information was removed or altered.
The dispute is not limited to books found on the open web. Google has long received or scanned publishing material through services such as Google Books and Google Play. The case may therefore examine whether access granted for search, preview, distribution or storage also permits model training.
Google had not publicly responded when the action was reported, so the allegations have not been tested in court.
Why it matters
AI companies often argue that training involves analysing works to learn statistical patterns rather than reproducing them as a conventional copy. Rights holders argue that large models could not be built without ingesting valuable creative material and that permission or compensation should therefore be required.
The source of the data may shape the legal analysis. A court could treat public web crawling differently from reusing files supplied under a commercial or platform relationship.
The bigger picture
The AI copyright debate is moving from broad theory into questions about contracts, data provenance and licensing boundaries. Model developers need better records of where training data came from and what rights were attached to it.
For publishers, the goal is not only damages. A favourable ruling could create leverage for licensing markets in which AI companies pay for high-quality, structured content. For Google, the case could affect how data collected across its many services can be used to improve AI products. The outcome may help define whether platform access is purpose-specific or can be extended to model training without a new agreement.