The Copyright Battle Continues: Tech Giant Nvidia Sued over AI Training Data
With the rise of artificial intelligence (AI), there have been an increasing number of lawsuits against big tech companies over the use of copyrighted content in training AI models.[1] One of these “tech giants” is Nvidia.[2] Nvidia is currently being sued by a group of authors—Brian Keene, Abdi Nazemian, and Stewart O’Nan—who allege that the company used their copyrighted works without permission to train its AI platform NeMo.[3] The authors’ works were included in a dataset of 196,640 books that were used to train NeMo to simulate ordinary written language.[4] The authors’ works at issue in the lawsuit include Nazemian’s Like a Love Story, Keene’s Ghost Walk, and O’Nan’s Last Night at the Lobster.[5]
The class action lawsuit, filed on Friday, March 8, 2024, claimed that Nvidia “admitted” it trained NeMo on the dataset of books, thereby infringing the authors’ copyrights.[6] The training dataset used by Nvidia was called “The Pile,” and one of the components making up “The Pile” is Books3—a collection of fiction and nonfiction books.[7] The lawsuit seeks damages for people in the U.S. whose copyrighted works were used to help train NeMo’s large language models (LLMs) in the last three years.[8]
What Is NVIDIA NeMo and Generative AI?
NeMo is an end-to-end platform created by Nvidia to develop custom generative AI.[9] Generative AI refers to deep-learning models that can generate content based on the data they were trained on.[10] Benefits of NeMo are that it can train and deploy generative AI anywhere, and quickly train, customize, and deploy LLMs.[11]
Challenges of Generative AI
Despite the advantages of generative AI, generative AI can lead to intellectual property law challenges.[12] Generative AI models can receive copyrighted information as part of the data used in training the model.[13] Using original works without a license to train generative AI models allows for unauthorized derivative works to be generated and, thus, copyright infringement.[14]
Ways to Mitigate Copyright Infringement Risks
Various steps can be taken to help prevent copyright infringement for generative AI models created by big tech companies. AI developers, such as Nvidia, should ensure that they comply with laws regarding the acquisition of data used to train their models.[15] To ensure compliance, AI developers should seek licenses from the individuals who own the works or intellectual property that the developers seek to add to their training data.[16] The individuals who own the intellectual property should also be compensated or share in the revenue generated by the AI model.[17] Consumers of AI models can help mitigate copyright infringement by asking the AI developers whether their models were trained with protected content.[18] Consumers can also avoid generative AI models in which the AI developers cannot confirm that their training data is properly licensed from the content creators.[19]
The Impact of Generative AI Models like NVIDIA NeMo
Generative AI will likely continue to profoundly impact society, such as with content creation, enabling many to do what only a few have the skills or advanced technology to accomplish.[20] However, companies developing generative AI models must respect the rights of those enabling content creation.[21]
Footnotes
