Ted Entertainment, Inc. v. Nvidia Corporation

On 26 November 2025, Ted Entertainment, Inc., Matt Fisher, and Golfholics, Inc. commenced class action proceedings against Nvidia Corporation in the United States District Court for the Northern District of California, alleging violations of the Digital Millennium Copyright Act. The Complaint asserts that Nvidia deliberately circumvented YouTube's technological protection measures to access and scrape millions of copyrighted videos, including theirs, contending that this conduct was undertaken to feed, train, and commercialise Cosmos, the defendant's foundational video world model designed to underpin its broader product ecosystem, including GeForce, Avatar, GR00T, and its autonomous vehicle systems.

According to the pleadings, Nvidia utilised three datasets composed entirely of YouTube video identifiers — HD-VG-130M (approximately 130 million clips derived from roughly 1.5 million videos), HDVILA-100M (approximately 100 million clips derived from roughly 3.1 million videos), and HowTo100M (over 100 million clips derived from roughly 1.2 million videos). The Complaint emphasises that these datasets contain only location pointers, not audiovisual files, meaning every clip required a separate act of downloading from YouTube and therefore a separate act of circumvention. It is further alleged that Nvidia deployed the open-source video-downloading programme yt-dlp combined with virtual machines that rotated IP addresses to evade YouTube's access protections and detection systems, enabling the defendant to harvest protected audiovisual content at scale without authorisation or compensation. Internal Slack communications cited in the Complaint, sourced from media reporting on leaked documents, allegedly reveal that Nvidia employees discussed building a "data factory" capable of yielding the equivalent of a human lifetime of visual experience in training data per day. Employees are also alleged to have raised concerns about the legality of the downloading, with the response being that it was an "executive decision" covered by "umbrella approval."

The Complaint further alleges that two of the three datasets were expressly licensed for academic or non-commercial use only, and that Nvidia was aware of these restrictions but used the datasets for commercial development of its generative AI model regardless. The claim also emphasises that the dataset creators did not own the underlying videos and therefore could not confer any right to download or commercially exploit the copyrighted works.

Claims advanced include circumvention of technological measures under the Digital Millennium Copyright Act, 17 U.S.C. § 1201(a). The Plaintiffs primarily seek a declaration that the Defendant wilfully circumvented the copyright protection systems of YouTube intended to protect Plaintiffs' and the Class Members' audiovisual content, followed by statutory damages (up to the maximum allowed per violation), injunctive relief, and attorneys' fees and costs