By now, it’s no secret to anyone that companies models have blatantly scraped and stolen online content, often copyrighted, to properly train their AI models. NVIDIA is now being sued for building its AI systems, allegedly with the help of YouTube videos.
A class action lawsuit has been filed against tech giant NVIDIA by YouTuber David Millette, who claims the company unjustly profited from his and other creators’ videos to build its AI systems. The lawsuit alleges NVIDIA violated California’s Unfair Competition Law and enriched itself at creators’ expense by scraping content from YouTube and other platforms without permission. The practice itself was unearthed following an investigation from 404 Media—Millette filed the lawsuit as one of the many affected parties.
The lawsuit is based on leaked documents and conversations revealing NVIDIA’s project, known as “Cosmos,” aimed at building a state-of-the-art video model. Employees reportedly scraped YouTube using virtual machines and discussed screen-recording Netflix and scraping Vimeo, raising ethical and legal concerns. While NVIDIA maintains it operates in compliance with the law, Millette’s complaint focuses on “unjust enrichment and unfair competition,” claiming damages exceeding $5 million. The lawsuit alleges NVIDIA unfairly profits from commercial products developed using unattributed reproductions of creators’ videos and ideas.
This lawsuit follows a similar one filed by Millette against OpenAI. By now, it’s pretty clear that almost every company building an AI model has trained its models, intentionally so or not, using copyrighted content. Will this result in a significant change in how AI companies source the data they train their models on? Unless these lawsuits actually go somewhere, it’s unlikely that this will result in an industry-wide change, or at least in the change that most of us want to see. One of the most important right now is the lawsuit against OpenAI by The New York Times, which was initially filed in 2023.
Source: 404 Media
