Nvidia sued for allegedly using Jamendo’s music to train AI without permission

Nvidia hit with US copyright suit claiming it trained AI models on Winamp subsidiary Jamendo’s music without permission

June 23, 2026 By Mandy Dalugdug

Jamendo , the music licensing platform owned by Belgium-headquartered Winamp Group , has filed a lawsuit in the United States against tech giant Nvidia.  Jamendo accuses the chipmaker of using copyrighted music data to train two of its AI audio models, Fugatto and Audio Flamingo, without authorization.

The move comes about a year after Jamendo publicly threatened legal action against both Nvidia and AI music company Suno over the alleged use of its catalog to train AI models.

But the complaint, a copy of which has been obtained and reviewed by MBW , names only Nvidia Corporation. It was filed on Monday (June 22) in the US District Court for the Northern District of California .

It alleges that Nvidia trained its Fugatto and Audio Flamingo models on the MTG-Jamendo Dataset , a research dataset built from Jamendo’s catalog and released for non-commercial use only.

Nvidia is the world’s most valuable company, with a market capitalization of approximately $5.1 trillion.

The MTG-Jamendo Dataset was created around 2019 through an academic collaboration with the Music Technology Group at Universitat Pompeu Fabra in Barcelona, and posted on GitHub for researchers.

The dataset contains “over 55,000 ” full audio tracks tagged with 195 tags spanning genre, instrument and mood/theme categories, drawn from music on Jamendo released under Creative Commons licenses.

The complaint brings six counts: two for copyright infringement, two for breach of contract, one for unjust enrichment, and one for unfair competition under California law.

Jamendo is seeking an injunction, plus actual damages and Nvidia ‘s profits of “no less than” €17.8 million (approx. $20.3 million ), or, in the alternative, statutory damages.

Those statutory damages run up to $30,000 per infringed work under the US Copyright Act , rising to $150,000 per work where infringement is found to be willful.

Jamendo alleges that Nvidia ‘s conduct was willful, which it says would entitle it to the enhanced figure.

Jamendo alleges that Nvidia ‘s own published research papers identify MTG-Jamendo as one of the datasets used to build Fugatto and Audio Flamingo .

Nvidia unveiled Fugatto, which stands for Foundational Generative Audio Transformer Opus 1, in November 2024 , describing it as “a Swiss Army knife for sound” .

According to the complaint, Nvidia has stated that its models were trained on “a large text and audio dataset with at least 20 million rows” that is “exclusively comprised of open source data.”

That phrase – “open source” – sits at the center of Jamendo ‘s dispute with Nvidia .

The dataset’s own documentation states that it is “made available solely for non-commercial research and academic use,” and that “any other use, including but not limited to commercial applications, requires prior written authorization from Jamendo S.A. ”

Jamendo ‘s argument is that “open source” does not mean free for any purpose.

> “As artificial intelligence continues to transform the music industry, we believe it is essential that creators and rights holders are properly recognized, compensated and protected.” Alexandre Saboundjian, Winamp Group

The dataset was offered for non-commercial research, Jamendo says, and training a commercial AI model falls outside that grant.

The terms attached to the dataset are layered, which is part of why the Nvidia question is contested.

The underlying tracks carry individual Creative Commons licenses set by each artist on Jamendo .

The dataset’s curated metadata, meanwhile, was released for non-commercial use, alongside the notice requiring Jamendo ‘s written authorization for any commercial use.

Whether AI training counts as “commercial use,” and whether a freely downloadable research dataset can be fenced off from it, are questions that will shape Jamendo ‘s case.

According to the complaint, when Jamendo invoiced Nvidia for the use, Nvidia responded that no contractual relationship had ever existed between the parties.

The heart of the US case is less the songs themselves than a ‘compilation’ copyright in the MTG-Jamendo Dataset.

Under US law, facts and raw data cannot be copyrighted, but an original selection, coordination, and arrangement of them can be – the principle set out by the Supreme Court in Feist Publications , Inc. v. Rural Telephone Service Co. in 1991 .

In Feist , the Supreme Court held that a phone book’s alphabetical white pages were too mechanical to qualify, ruling that copyright rewards a “modicum of creativity,” not the “sweat of the brow” of mere effort.

Jamendo says its dataset clears that bar through its curation and tagging, and that the US Copyright Office issued Registration No. TX-9-606-566 for the compilation on June 17, 2026 .

Its claim, in short, is that Nvidia copied not just uncopyrightable facts but the protected structure Jamendo built around them.

The Jamendo dispute has also reached a court in Belgium .

In a press release issued on Tuesday, Winamp Group said that on June 11, 2026, the Ghent Enterprise Court confirmed its jurisdiction in the proceedings initiated by Jamendo against NVIDIA Technologies Belgium concerning a commercial claim of approximately EUR €16 million.

“This claim arises from Jamendo’s invoicing of the alleged unauthorized use of musical content (more than 55,000 works) and related data exploited by Jamendo,” the company said in the press release, adding: “The underlying facts of this commercial claim are also at the heart of the proceedings initiated in the United States under intellectual property and contractual legal theories”.

WINAMP said that “through this decision, the Court rejected the procedural objections raised by NVIDIA Technologies Belgium and established a timetable for the continuation of the proceedings”.

The parties are expected to exchange written submissions between the end of 2026 and the beginning of 2027, with oral pleadings currently scheduled for June 24, 2027.

Commenting on the legal action in the US and in Belgium, Alexandre Saboundjian, Chief Executive Officer of Winamp Group, said: “These actions reflect our commitment to protecting the rights of Jamendo and the artists who entrust us with the commercialization of their works.”

“As artificial intelligence continues to transform the music industry, we believe it is essential that creators and rights holders are properly recognized, compensated and protected.”

The legal action  follows Jamendo  publicly threatening  legal action against both Nvidia and AI music company Suno last year. The Winamp Group has not said publicly whether separate action against Suno will follow.

In a statement last year, Alexandre Saboundjian , CEO of Jamendo and Winamp , said: “Our music catalog is not free for exploitation by commercial entities building AI models without permission or compensation.”

“ Nvidia and Suno ‘s use of our artists’ work without authorization is not only unlawful, it is a direct threat to the livelihoods of independent musicians worldwide. We will not stand idly by.”

Suno , for its part, remains in litigation with Universal Music Group and Sony Music and faces suits from European rights bodies, having settled with Warner Music Group in November 2025 .

The company argues that training on copyrighted music is permitted as fair use, and in June 2026 reached a $5.4 billion valuation in a new funding round. Music Business Worldwide

News United States Audio Flamingo Fugatto Jamendo Nvidia Suno Winamp Group

Related Posts

Four music datasets holding millions of tracks are being shared among AI developers, The Atlantic reports

_Originally reported by [Music Business Worldwide](https://www.musicbusinessworldwide.com/nvidia-hit-with-us-copyright-suit-claiming-it-trained-ai-models-on-winamp-subsidiary-jamendos-music-without-permission/)._