Leaked documents and internal communications obtained by 404 Media reveal that Nvidia has been scraping videos from YouTube and other sources to amass training data for its AI products.
Nvidia’s internal Slack chats, emails, and documents show that the company has been collecting videos to enhance its AI models, including the Omniverse 3D world generator, self-driving car systems, and “digital human” products.
The project, codenamed Cosmos, has not yet been publicly released. Despite questions about the legality of this data collection, Nvidia maintains that its practices comply with copyright laws.
Nvidia’s extensive video scraping strategy involved downloading videos from platforms like Netflix and YouTube. Employees used tools such as the open-source YouTube downloader yt-dlp, coupled with virtual machines that regularly refreshed IP addresses to avoid detection.
According to internal emails, the company utilized 20 to 30 virtual machines on Amazon Web Services to download an estimated 80 years’ worth of video content daily.
Ming-Yu Liu, Nvidia’s vice president of Research and a leader on the Cosmos project, emphasized in a May email that the goal was to create a comprehensive video foundation model. Despite Nvidia’s claims of compliance, the use of copyrighted content without explicit permission has raised significant concerns.
A former Nvidia employee, speaking anonymously to 404 Media, confirmed that the video scraping included content from Netflix, YouTube, and other sources.
The documents also show that Nvidia’s leadership was aware of potential legal issues. Emails reveal that project managers assured employees that they had the necessary clearance from higher-ups.
Nvidia’s spokesperson defended the company’s practices, stating that copyright law allows for the use of facts, ideas, data, or information for transformative purposes, such as AI model training.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
