A 3D Artist Is Suing Meta, Nvidia, and Roblox Simultaneously Over AI Training Data. It’s the Biggest Case of Its Kind

Giving something away freely and then seeing someone else make large profits from it can be particularly frustrating. Los Angeles-based digital artist Austin Beaulier appears to have reached that stage prior to March 26, 2026, as he filed three proposed class action lawsuits in California federal court on that Thursday. Roblox, Nvidia Corporation, and Meta Platforms are the defendants. all at once. It’s the kind of legal action that causes you to take notice.

Beaulier works in the field of 3D modeling, which is the laborious process of creating digital objects that appear in animated environments, virtual worlds, and video games. He had been sharing his models via public repositories, which are the kind of shared databases that are essential for open communication and collaboration within the creative community. The terms, which are governed by Creative Commons licenses that allow downloading and use but firmly forbid commercial exploitation, were fairly clear. Those licenses are specifically intended to stop someone from making money off of your free model.

According to the complaints, Roblox, Nvidia, and Meta went too far. Beaulier claims that in order to train their generative AI systems, each company extracted millions of 3D models from these repositories. The specific tools mentioned in the filings are Roblox’s Cube3D, Nvidia’s TRELLIS-500K, and Meta’s SAM-3D. These aren’t lab-sitting prototypes for specialized research. These are production-grade systems used to create 3D assets for commercial products, such as animation pipelines, virtual environments, and games. In essence, the claim is that Big Tech utilized the open creative commons as a free resource and fed it into machines that are now in direct competition with the artists who initially populated the commons.

How each of the three businesses will react is still unknown. As of the filing date, none of them had commented, nor had Beaulier’s lawyer. However, dozens of other AI copyright cases that have accumulated in courts over the past few years have already made the legal framework they’ll probably rely on familiar. The industry’s go-to defense has been the fair use argument, which holds that AI systems transform source material into something truly new. Courts haven’t fully addressed the question of whether it holds up when millions of 3D objects are scraped without permission.

The scope of this lawsuit is what gives it a unique feel. Image generators have been sued by visual artists. AI writing tools have been sued by authors. Language models trained on news organizations’ archives have been targeted. However, it is unprecedented to file three class action lawsuits at the same time against three different large corporations in the same category of creative work. Beaulier is looking for more than personal solace. He is requesting that the court allow him to speak on behalf of a class of 3D artists across the country, possibly thousands of people whose work may have been included in these training datasets without their knowledge or consent.

In the broader AI copyright discussion, which has tended to concentrate on authors, illustrators, and photographers, there is a feeling that the 3D art community has been somewhat neglected. The people who create the environments found in video games and virtual worlds occupy an odd middle ground because, once implemented, their work is ubiquitous and invisible in the sense that most users never consider who created the items that fill a virtual space. Beaulier’s repositories were based on a mutual understanding that unrestricted commercial use does not equate to free distribution. If the accusations are true, there was a covert, large-scale breach of that trust.

It’s important to keep in mind that AI training was not considered when designing Creative Commons licenses. They were designed to be used in a world where people would remix, adapt, and build upon artistic creations. There has never been a direct test of whether feeding millions of files into a neural network qualifies as the type of “commercial exploitation” that these licenses prohibit. Every creator who has ever contributed to a shared repository under the presumption that “free to use” had limitations will be interested in the decision made by the courts.

It’s difficult to ignore the fact that the legal system is still far behind the speed of the technology it is being asked to regulate as this case develops over the ensuing months. The businesses can litigate slowly because they have the means to do so. This isn’t merely a symbolic move, as Beaulier has Milberg, a firm with actual class action experience. Many 3D artists will be closely watching to see if the work they shared in good faith was protected in any way. The cases have been filed, and the court records are available to the public.