Unauthorized books in AI training: authors file lawsuit against Nvidia

San Francisco: Chipmaker Nvidia is facing legal action from three authors - Brian Keene, Abdi Nazemian, and Stewart O'Nan - over claims that their copyrighted books were used without permission to train the company's NeMo AI platform.

The authors argue that their works, part of a dataset of approximately 196,640 books, were utilized to train NeMo's large language models, simulating ordinary written language.

Nvidia allegedly admitted to training NeMo on the dataset, leading to its takedown in October due to reported copyright infringement, according to the authors.

Filed as a proposed class action in a San Francisco federal court on Friday, the lawsuit seeks unspecified damages for individuals in the United States whose copyrighted works contributed to training NeMo's large language models over the past three years. The authors mentioned in the lawsuit include Brian Keene's 2008 novel "Ghost Walk," Abdi Nazemian's 2019 novel "Like a Love Story," and Stewart O'Nan's 2007 novella "Last Night at the Lobster."

Nvidia, a key player in artificial intelligence with its chips powering AI applications, declined to comment on the lawsuit.

This legal action follows a trend of litigation against companies involved in generative AI, where AI platforms generate new content based on various inputs. Nvidia's NeMo platform has been promoted as a fast and affordable solution for adopting generative AI technology.

This lawsuit adds Nvidia to the list of companies facing legal challenges related to the use of generative AI, including OpenAI and Microsoft. The legal dispute underscores the increasing scrutiny on AI's impact on intellectual property and copyright concerns.

Nvidia's stock has seen significant growth, rising nearly 600% since the end of 2022, resulting in a market value of nearly $2.2 trillion for the Santa Clara, California-based chipmaker.