Photo Credit: Brecht Corbeel
Another day, another high-stakes AI lawsuit: Now, “independent online music platform” Jamendo is suing Nvidia for copyright infringement and breach of contract for allegedly training on tens of thousands of tracks without authorization.
Luxembourg-based Jamendo submitted the straightforward complaint to a California federal court yesterday, roughly eight months after taking legal action against Nvidia in Belgium.
As described by the stateside suit, the older case is unfolding in Ghent, where a judge earlier in June “rejected Nvidia Belgium’s jurisdictional objection, ordering that the matter shall proceed and be heard on its merits.”
But with Nvidia Belgium still disputing “the validity and enforceability of Jamendo’s invoice” – more here in a moment – Winamp Group-owned Jamendo evidently believes that the time’s right to seek relief in the States as well.
According to the plaintiff, Nvidia unlawfully trained its Fugatto and Audio Flamingo models on a music database created by Jamendo and the Music Technology Group (MTG) at Barcelona’s Universitat Pompeu Fabra.
Though Jamendo hosts hundreds of thousands of songs in total, the collaborative MTG database contains about 56,000 tracks (besides detailed metadata), per the suit.
And as the subset is publicly “available solely for non-commercial research and academic use,” any commercial use would require Jamendo’s express approval.
Unsurprisingly, then, that “Nvidia’s publicly-available publications expressly identify MTG-Jamendo as one of the datasets Nvidia utilized for its Fugatto and Audio Flamingo products” isn’t sitting right with the plaintiff.
It probably doesn’t need saying that this alleged use was unauthorized. And shortly after learning of the alleged training free-for-all in March 2024, “Jamendo sought to initiate discussions with Nvidia regarding a commercial license,” according to the suit.
Another point that goes without saying: “multiple efforts at negotiation with Nvidia and its representatives,” running through June 2025, failed to bring about the desired result for Jamendo.
In response, the entity went ahead and forwarded Nvidia the aforesaid training invoice and took the previously noted legal action in Belgium.
Included as an exhibit with the stateside complaint, this invoice demanded €289 (currently $329) a pop for the use of the relevant database’s 55,600 tracks – or €16.1 million/$18.3 million in total for an international license.
“Nvidia has failed to cure its default, has not paid the invoiced amount, and has continued to retain the benefits of its unauthorized use of the MTG-Jamendo Dataset,” the suit reads.
All told, Jamendo is suing for straight infringement, breach of contract (concerning the allegedly unlawful commercial use of the IP and the alleged violation of Jamendo’s terms), unjust enrichment (“by using the MTG-Jamendo dataset, Nvidia obtained substantial benefits”), and unfair competition.
“Upon information and belief, by avoiding the costs of licensing or independently developing a comparable dataset, Nvidia has reduced its research and development expenses and accelerated the development of its AI systems,” the latter claim maintains.
On the relief front, Jamendo is seeking, among other things, a minimum of €17.8 million/$20.3 million – or the invoice total plus interest. DMN reached out to Nvidia for comment but didn’t receive a response in time for publishing.
