Nvidia Sued By Authors For Training AI With Copyrighted Works

Graphics chip maker Nvidia is facing a class action lawsuit from three authors who say the company illegally used their works and those of others to train its NeMo Megatron AI model, the latest lawsuit of its kind to target an AI company

Databricks, the parent company of MosaicML, was also hit by a parallel lawsuit, with both filed in San Francisco federal court.

Both lawsuits claim the respective companies used a library of pirated ebooks called Books3 to train their large language models (LLMs).

Similar lawsuits have been filed against AI companies including OpenAI, its main investor Microsoft, Meta and others by authors and by The New York Times for training their AI systems on copyrighted content, which they argue violates copyright law.

Pirated books

Books3 is a dataset of about 196,640 books, mostly by contemporary authors such as Stephen King and John Grisham.

The three authors, Abdi Nazemian, Brian Keene, and Stewart O’Nan, said the “model cards” attached to each Nvidia AI model say they were trained on the Books3 dataset.

“We respect the rights of all content creators and believe we created NeMo in full compliance with copyright law,” Nvidia said in a statement.

Databricks recentluy purchased AI company MosaicML, which makes the MPT series of AI models.

Copyright questions

The complaint against Databricks similarly says that public information about MPT shows those models were also trained on Books3 content.

The lawsuits name works including Keene’s 2008 novel Ghost Walk, Nazemian’s 2019 novel Like a Love Story and O’Nan’s 2007 novella Last Night at the Lobster.

Nvidia dominates the market for AI chips and launched its own NeMo models in 2022.

The company has seen its stock price soar in recent months due to high demand for its AI hardware, driving its market capitalisation over $2 trillion (£1.6bn).