Google faces a new class-action lawsuit in the United States, after a group of major publishers and authors accused it of using copyrighted books and works to train the Gemini artificial intelligence model without obtaining prior permission.

The list of plaintiffs includes prominent publishers such as "Hachette", "Cengage", and "Elsevier", along with American author Scott Turow and the organization S.C.R.I.B.E., who demand that the company be held accountable for what they describe as widespread infringement of intellectual property rights.

Allegations of hiding copyright information

The lawsuit not only accuses Google of using protected content, but also alleges that the company removed or altered copyright information associated with these works, in order to conceal the fact that Gemini models were trained on materials 'obtained illegally,' according to the lawsuit.

This case is the latest in a series of lawsuits facing AI companies such as Google, OpenAI, Meta, and Anthropic, due to their use of copyrighted works to train their models.

Previous rulings supporting AI companies

Despite most of these cases still pending in courts, two courts in California recently issued preliminary rulings that considered the use of copyrighted materials in training AI models to fall under the 'Fair Use' principle according to US copyright law, a law that predates the internet and has not been updated to keep pace with AI developments.

But these rulings do not mean that all cases are settled, as each lawsuit differs according to its circumstances and the evidence presented.

Legal precedent against Anthropic

In contrast, the sector saw a notable development when Anthropic was forced to pay $1.5 billion for using pirated works to train its models, in one of the largest settlements in the history of US copyright cases.

The settlement allowed approximately 500,000 writers to receive compensation of at least $3,000 each, while a number of authors preferred not to join the settlement in order to continue pursuing AI companies legally.

Google Books at the heart of the case

The new lawsuit is based on the long relationship between Google and the publishing industry, as the plaintiffs explained that they previously allowed the company to use their books within the Google Books service, which allows searching inside books and displaying short excerpts with metadata, without showing the full text.

But they emphasize that this permission was limited to the search service only, and did not include using those books to train AI models.

The lawsuit alleges that Google used copies of books from Google Books, as well as books published through Google Play Books, to train Gemini without obtaining permission from rights holders.

The lawsuit states: "Google illegally copied the works within these limited-use programs to train AI, despite knowing that it did not have permission to do so."

Internal document raises suspicions

The plaintiffs also cited an internal Google document, which they say shows that the company was aware of the legal risks of using copyrighted books to train AI models.

According to the lawsuit, the document warned that this practice could be 'highly problematic for Google,' and could expose it to fines ranging from $10 billion to $100 billion if rulings are issued against it.

The case opens a new chapter in the escalating conflict between AI companies and the publishing sector, amid ongoing legal debate over the limits of 'fair use' in the age of generative models.

Advertisement

Advertisement