Skip to content

Analytics cookies

We use Google Analytics and Google Tag Manager to count visits and see which pages are useful. If you accept, Google Analytics may set cookies to recognise your return visits. If you decline, it sets no cookies, but Google still receives a cookieless signal for each page you view. We don’t send Google your name, email address or registration details, and you can change your choice at any time under “Cookie settings” at the bottom of the page. Privacy

Munich court rules Suno infringed GEMA copyrights by training its music AI on six songs

Jul 31, 2026 · Gathered automatically from public sources. Not reviewed by a person.

What happened

On 31 July 2026, the Munich Regional Court I (Landgericht München I) ruled largely in favour of GEMA, the German music collecting society, against Suno Inc., the US provider of an AI music generator, in case no. 42 O 763/25. The 42nd Civil Chamber, presided over by Judge Elke Schwager, heard the case in March 2026; GEMA had filed the action in January 2025. This was GEMA's second copyright case against a generative-AI provider, after its November 2025 win against OpenAI over song lyrics, which is under appeal.

The works and the evidence

The case concerned six musical compositions in GEMA's repertoire: "Daddy Cool" and "Rasputin" (Frank Farian), "Forever Young" and "Big in Japan" (Marian Gold, Bernhard Lloyd and Frank Mertens), "Atemlos" (Kristina Bach), and "Mambo No. 5" (David Lubega and Christian Pletschacher). Song lyrics were not at issue; the musical compositions were. It was undisputed that Suno's models were trained using the six works, which Suno extracted from YouTube by "stream-ripping", circumventing YouTube's "rolling cipher" technical protection measure. GEMA showed that simple prompts naming a title, lyrics and musical style produced audio that corresponded in melody, harmony and rhythm to the originals.

What the court held

The court found the works were effectively stored in Suno's models (reportedly versions 3.5 and 4) on servers in Germany through "memorisation" — retention of parts of the original works in reproducible form — and that the similarities were too extensive and specific to be coincidental. It held this was unauthorised reproduction under German copyright law, that the text-and-data-mining exception did not apply because the works were retained rather than merely analysed, and that Suno, not its users, was responsible for infringing outputs because users entered only basic prompts while Suno designed, trained and operated the system. For training carried out in the US, the court applied US copyright law and rejected Suno's fair-use defence, distinguishing the US decisions in Bartz v. Anthropic and Kadrey v. Meta on the basis that there the outputs did not closely reproduce the training works.

Orders and status

The court ordered Suno to cease the unauthorised reproduction of the works and stop using them to train its models, to disclose information about revenue connected with the infringements, and to pay damages, the amount of which has not yet been determined. The judgment is a first-instance ruling: it is not yet final or enforceable, and Suno has said it intends to explore all options, including an appeal. These are the court's findings at first instance, not a final determination.

Sources