Copyright claims against generative artificial intelligence are getting harder to keep up with a new proposed class action that hit the courts last week. This time, the authors are suing NVIDIA for it AI platform NeMoa language model that allows businesses to build and train their own chatbots, Ars Technica informed. They claim that the company trained this data on a controversial dataset that illegally used their books without consent.
Authors Abdi Nazemian, Brian Keene, and Stewart O’Nan demanded a jury trial and asked Nvidia to pay damages and destroy all copies of the Books3 dataset used to power the NeMo large language models (LLM). They claim the dataset copies a shadow library called Bibliotek, which contains 196,640 pirated books.
“In general, NVIDIA admitted to training NeMo Megatron models on a copy of The Pile data set,” the lawsuit states. “Therefore, NVIDIA must have trained the NeMo Megatron models on a copy of Books3 because Books3 is part of The Pile. Some books written by plaintiffs are part of Books3, including infringing works – and thus NVIDIA must train NeMo Megatron models of one or more copies of the Infringing Works, thereby directly infringing Plaintiffs’ copyrights.
In response, NVIDIA said The Wall Street Journal “we respect the rights of all content creators and believe we created NeMo in full compliance with copyright laws.”
Last year, OpenAI and Microsoft hit it off copyright claim from non-fiction authors, the companies claimed that they made money from their works but refused to pay them. There was a similar claim has been activated earlier this year. One on top of that claim from news organizations such as The Intercept and Raw Storyand of course the legal action that started it all The New York Times.
