Jamendo alleges Nvidia AI music model trained on dataset intended for academic research

A lawsuit filed by the music platform Jamendo alleges that Nvidia’s AI music system trained on a free dataset that was intended only for academic research, drawing claims that the company used copyrighted songs without permission. The case, reported June 23 by Billboard, centers on how Nvidia’s “Fugatto” music generator was trained and whether the training data and process complied with the dataset’s stated purpose and rights holders’ permissions.

According to the reporting, Jamendo argues that Nvidia’s use of the dataset went beyond what the dataset was meant to support. The complaint characterizes the alleged training as a violation of terms tied to the dataset’s distribution, and it frames the conduct as improper use of creative works rather than a narrowly scoped research effort. Jamendo’s position is that training a commercial-grade generative system on material provided under academic-only intentions is not protected by the dataset’s original framing.

The dispute has turned on the relationship between what dataset providers describe as permissible use and how AI developers later apply that data. In Jamendo’s account, the dataset in question was “intended only for academic research,” while Nvidia’s music generator is being treated as a broader application. Billboard’s report quotes Jamendo’s description of the conduct as “a blatant violation,” reflecting the platform’s view that the training was not authorized by the original dataset terms.

Nvidia, Jamendo, and any other parties involved are also likely to address what legal standards govern AI training and the use of copyrighted music in generative models. The filing highlights that the controversy is not only about whether songs were included in training, but also about the contractual or licensing boundaries, if any, set by how the dataset was made available. The lawsuit therefore puts emphasis on intended use, permissions, and whether the downstream model’s function alters the rights analysis.

If the court finds that training fell outside permission or violated licensing conditions, the remedies sought could include damages and injunctive relief limiting use of the model or training materials, depending on what Jamendo requests. The case also carries practical stakes for music licensing norms, because many dataset and training workflows depend on assumptions about permissible purposes and downstream deployment.

For artists and music distributors, the litigation underscores a growing pressure point in the AI music sector: the gap between dataset descriptions and real-world deployment. If Jamendo’s allegations are sustained, rights holders and platforms may push for tighter controls around how music is collected, tagged, and licensed for machine learning, including clearer limits on the types of models or uses allowed.

As of publication of the Billboard report, the lawsuit’s next steps would depend on procedural developments, including service, responses from Nvidia, and the court’s consideration of motions that can address licensing terms, authorization, and other legal thresholds. The case is also expected to generate further attention on how AI developers document training data sources and comply with stated conditions tied to music libraries distributed for specific purposes.