Nvidia Faces Class Action Lawsuit Over Unlawful Use of YouTube Videos for AI Training
A class action lawsuit has been filed against Nvidia Corporation, alleging the unlawful use of YouTube videos to train its “Cosmos” AI model. The complaint, filed on August 14, 2024, by Plaintiff David Millette on behalf of other YouTube users, raises serious concerns about consent and unfair business practices.
Millette filed a similar suit against OpenAI last month. The analysis is very similar, so we’ll focus on Nvidia here.
Background on the Lawsuit
The suit names David Millette as the primary Plaintiff. Millette has been uploading videos to YouTube since June 2009, discovered that Nvidia . The lawsuit claims that Nvidia scraped these videos and used Millette’s content to train its Cosmos AI, a deep learning service that supports various Nvidia products without Millette’s consent.
Millette cites several sources (that are now unavailable) describing internal communications discussing compiling 38.5 million URLS and scraping 100,000 YouTube videos, essentially a lifetime's worth of data everyday, including MIllette’s videos.
Millette is pursuing two causes of action: unjust enrichment and unfair competition.
Unjust Enrichment
Millette and class members were not compensated or credited for contributing their content to Nvidia’s Cosmos project. By using Plaintiff’s videos without permission to develop and enhance its AI products, Nvidia exploited their content for commercial gain.
Unjust enrichment has its origins in common law, dating back to the middle ages. Under U.S. law, a claim for unjust enrichment requires the plaintiff to prove:
The defendant was enriched
The defendant was enriched at the plaintiff's expense
It's against equity and good conscience to allow the defendant to keep what the plaintiff is seeking to recover
The benefit given to the defendant must not be a gift, forced upon them without reason, or something too vague to measure. It also can't come from someone interfering in the defendant's business without a good reason.
Millette and class members were not compensated or credited for contributing their content to Nvidia’s Cosmos project. The complaint asserts that by using Plaintiff’s videos without permission to develop and enhance its AI products, Nvidia exploited their content for commercial gain.
There are problems with this argument, primarily in the second factor: Was Nvidia enriched at Millette’s expense? The average YouTuber receives between $0.01 and $0.03 per view. Assuming Nvidia’s scraping requires 1 view of Millette’s videos and Nvidia used a scraping method that circumvented triggering a view in YouTube’s, Millette lost $0.01 and $0.03 per video. Assuming he’s posted 100 videos, actual damages are likely between $1 and $3. That’s not likely enough to support an unjust enrichment claim, since allowing Nvidia to keep $3 is not unconscionable.
Unfair Competition
The unfair competition claim is a longshot. Unfair competition includes a variety of practices, such as:
Making false allegations about a competitor
Misleading the public about the quality or characteristics of a product or service
Using physical force or coercion to sell goods or services
Violating non-compete, non-solicitation, or non-disclosure agreements
Intentionally interfering with prospective contractual relationships
None of the facts presented in the complaint describe these practices. Rather the complaint states that Nvidia engaged in “unfair business practices [that] violate the UCL because they are unfair, immoral, unethical, oppressive, unscrupulous, or injurious to consumers.” The closest this list of synonyms for unfair comes to actually presenting a case for unfair competition is “injurious to consumers,” which is not supported by any of the facts presented in the complaint.
What about YouTube’s Terms of Service?
YouTube has made it abundantly clear in its terms of service that scraping is not allowed. Nvidia undoubtedly violated YouTube’s terms of service, scraping 100,000 videos per day.
Does this breach constitute an unfair business practice? Probably.
Can creators whose content is hosted on YouTube raise a cause of action based on Nvidia's breach of YouTube’s terms of service? Maybe?
Is YouTube liable to its content creators for allowing Nvidia, OpenAI, and other companies training AI models? Perhaps?
This isn’t exactly uncharted territory. There is a body of case law that came out of hosting services and website owners suing Google for indexing their websites. But, there is a fundamental difference between identifying where content can be found (Google’s indexing) and using the content to create more content (training AI models). As a legal commentator, I call this job security.
Doesn’t YouTube have a Claim against Nvidia?
Did YouTube give Nvidia permission to scrape it’s content? If not, why didn’t YouTube stand up for itself and its content creators and either stop Nvidia’s scraping or pursue legal action? It is reasonable to assume that YouTube suffered actual injury from Nvidia’s scraping activities. Mass scraping can cause a variety of problems for companies hosting content. However, the cost of litigation and threat of discovery may have dissuaded YouTube from pursuing Nvidia.
Conclusion
To be fair, Millette raises important questions about scraping user content for training data, consent, and content ownership. Unfortunately, the claims of unjust enrichment and unfair competition are legally shaky, particularly due to the minimal financial harm suffered by individual YouTubers like Millette. The real injured party here may be YouTube itself, given Nvidia's likely violation of its terms of service. However, it's unlikely that YouTube will take action, as the costs and risks of litigation outweigh the potential benefits. In the end, while this case raises valid concerns about AI training practices, it’s doubtful that it will bring meaningful change unless larger platforms like YouTube step in to protect their creators' rights.
Tags: Nvidia, Class Action Lawsuit, AI Training, YouTube, Data Privacy, Legal News
