Jamendo has sued Nvidia again. The Winamp-owned music licensing platform filed a new copyright complaint against Nvidia in the Northern District of California on October 12, 2026, case 5:26-cv-11754, days after it voluntarily dismissed a near-identical suit filed on June 22, according to AI Musicpreneur and IP Law Daily coverage dated October 14.
The core claim has not moved. Jamendo says Nvidia trained its Fugatto and Audio Flamingo audio models on the MTG-Jamendo dataset, which is licensed for non-commercial research only, and it wants no less than 17.8 million euros.
What is the MTG-Jamendo dataset?
The dataset was built around 2019 with the Music Technology Group at Universitat Pompeu Fabra in Barcelona. AI Musicpreneur reports it holds more than 55,000 full tracks from about 3,500 artists, tagged by genre, instrument, and mood, and sits on GitHub under a Creative Commons BY-NC 4.0 license. The “NC” means non-commercial. The repository adds that the dataset is for non-commercial research and academic use only, and that anything else needs Jamendo's written authorization.
Jamendo says Nvidia's own research papers name MTG-Jamendo as a training source for Fugatto and Audio Flamingo. Nvidia has described its training data as “exclusively comprised of open source data,” according to the reporting. The complaint answers that a license restricting commercial use is not open source under the Open Source Initiative's definition.
IP Law Daily summarizes the complaint as alleging copyright infringement and related California state-law claims for unauthorized use of Jamendo's library to train those models, seeking injunctive relief and monetary damages, including statutory fees. AI Musicpreneur says the new complaint carries six counts: two for copyright infringement, two for breach of contract, one for unjust enrichment, and one for unfair competition. One copyright count covers the dataset as a compilation; the other covers individual songs and recordings whose artists gave Jamendo exclusive commercial licensing rights.
Why file again, and what else is pending?
Jamendo has not said publicly why it dismissed the June case and refiled instead of amending. The new case starts from the beginning under a new docket number. AI Musicpreneur notes Jamendo already walked away from one AI case in 2026: in August it dropped a lawsuit against Suno that also claimed training on research-only Jamendo data.
The US filing is the second front. Jamendo says it learned in March 2024 that Nvidia had used its data, tried to negotiate a license through June 2025, then invoiced Nvidia and Nvidia Technologies Belgium on September 29, 2025. A formal notice on October 14, 2025 demanded about 17.8 million euros, including default interest and a lump-sum indemnity. On June 11, 2026, a court in Ghent rejected Nvidia Belgium's objection to jurisdiction and ordered that case heard on its merits, according to AI Musicpreneur. Nvidia Belgium still disputes the invoice.
Why does a research-only license matter to working artists?
Most independent catalogs never appear in a named academic dataset. Jamendo's fight is still a warning about how “open” music ends up inside commercial AI stacks. If a Creative Commons non-commercial tag does not bind a model trainer, the practical protection for artists who posted tracks for research or free non-commercial use collapses. That sits beside Sony's campaign to pull hundreds of thousands of deepfakes, the consent fight over voice and likeness, and streaming platforms' new AI labels, including Deezer's search toggle. The Nvidia docket will test whether “research only” on a music zip file means anything once the weights ship in a commercial product.