Publishers and authors initiate class action against Google over Gemini AI training

A coalition of major book publishers and prominent authors has filed a class action lawsuit in New York, accusing Google of illegally utilising millions of copyrighted works to train its Gemini artificial intelligence model. The plaintiffs, comprising Hachette Book Group, Cengage Learning, Elsevier, author Scott Turow, and his publishing entity S.C.R.I.B.E, argue that the tech giant breached copyright laws by repurposing content provided for limited purposes to generate material that directly competes with human creators.

The complaint, lodged on Tuesday, contends that Google “secretly copied” millions of works from services such as Google Books. The plaintiffs allege that this data was subsequently used to train Gemini, an AI system capable of generating content that mimics the expressive elements and creative choices of specific authors. The suit characterises the scale and speed at which Gemini can produce books as unprecedented, posing a direct threat to the commercial viability of human writers.

Legal representatives for the plaintiffs are seeking an injunction to halt the alleged infringing activities, alongside unspecified damages. The filing requests class action status, aiming to consolidate the grievances of multiple rights holders into a single legal proceeding. The core of the argument rests on the assertion that the AI-generated outputs do not merely reference existing works but actively replicate the creative decisions of the original authors, thereby undermining their market position.

This litigation represents the latest escalation in the ongoing conflict between the creative industries and artificial intelligence developers regarding copyright infringement. The same group of publishers and author Turow previously filed a lawsuit against Meta in May on similar grounds in a New York court. These actions occur against a backdrop of evolving judicial interpretations of “fair use” in the context of AI training data.

Recent legal precedents have offered mixed outcomes for AI developers. In September, a US judge approved a $1.5 billion settlement between Anthropic and several authors, ruling that the use of books to train the Claude model constituted “fair use” under US law, although other uses of pirated materials were found to be infringing. Similarly, Meta secured a partial victory last year when a judge ruled that its use of copyrighted materials was transformative and thus fell under fair use protections. The outcome of the current case against Google remains undetermined, with the legal merits of the “fair use” defence yet to be adjudicated in this specific instance.