GEMA has won its legal battle against the AI music service Suno. On July 31, 2026, the Munich I Regional Court ruled that Suno infringes copyrights when the company trains its AI using copyrighted songs and subsequently plays them back in its output. This is the first ruling in Europe regarding AI-generated music —and the message to the industry is clear: no training without a license.
Key Points at a Glance
- The Munich I Regional Court ruled in favor of GEMA on July 31, 2026 (Case No. 42 O 763/25).
- Suno may not reproduce six copyrighted works, such as “Atemlos” or “Mambo No. 5,” without a license or use them for AI training.
- The court considers the so-called “memorization” of the songs in the model to be reproduction under Section 16 of the German Copyright Act (UrhG).
- GEMA was granted injunctive relief , the right to information, and damages.
- The ruling is not yet final —Suno may file an appeal.
What was the dispute between GEMA and Suno about?
Suno is a U.S. company whose AI tool generates finished, playable songs from simple text prompts. This is precisely what proved to be the provider’s undoing: GEMA was able to demonstrate in court that simple inputs can generate audio content that sounds strikingly similar to world-famous originals. The prompts specified only song titles, lyrics, and musical style—no specifications regarding melody, harmony, rhythm, or arrangement.
GEMA had already filed a lawsuit on January 21, 2025, after Suno failed to respond to a request for licensing. At the heart of the case were six compositions from the GEMA repertoire.
| Song in Question | Known for |
|---|---|
| Atemlos (Through the Night) | Helene Fischer (Composition: Kristina Bach) |
| Forever Young | Alphaville |
| Big in Japan | Alphaville |
| Mambo No. 5 | Lou Bega |
| Rasputin | Boney M. |
| Daddy Cool | Boney M. / Frank Farian |
The GEMA-Suno Ruling: AI Training Without a License Is Illegal
The 42nd Civil Chamber of the Munich I Regional Court, presided over by Judge Elke Schwager, largely ruled in GEMA’s favor. Suno may not reproduce the six copyrighted works without the rights holders’ consent—and this expressly includes their use as training material for AI. In addition, the court granted GEMA’s claims for disclosure and damages. Suno must therefore disclose the revenue generated from the disputed uses.
For GEMA, this is a landmark victory: The court confirmed that the systematic use of the repertoire and its commercial exploitation are subject to a licensing requirement. Suno charges a subscription fee for the premium version of its tool, thus pursuing a clearly profit-oriented business model.
Memorization: Why the Court Refers to “Reproduction”
The most interesting part of the ruling lies in the reasoning. Suno argued that the training data was not actually contained in the model—only learned patterns and generalizable features were stored. The court did not follow this line of reasoning.
In the court’s view, memorization took place during the training: The works are reproducibly contained on servers located in Germany and can be reproduced in the output. Given the complexity and length of the musical pieces, it could not be a coincidence that they appear in the output. This constitutes a reproduction under Section 16 of the German Copyright Act (UrhG), and the recognizable reproduction in Germany constitutes a public performance under Section 15(2) of the UrhG.
The court also attributed liability to Suno: the architecture and memorization fall within the operator’s sphere of control, and the operator thereby helps determine the content of the output. The fact that GEMA had entered the prompts itself does not change this—the prompts were kept simple and open-ended.
Also of interest is the method Suno used to obtain its training material: According to the court proceedings, the company used stream-ripping techniques to extract music from YouTube, thereby circumventing the rolling cipher—that is, precisely the technical protection measure intended to prevent downloading. Suno could not invoke the text-and-data-mining exception under Section 44b of the German Copyright Act (UrhG): The memorization of entire works goes beyond the permitted purposes of analysis.
Why a German Court Ruled on U.S. Training
Suno trained its models in the U.S.—yet the Munich Regional Court I (LG München I) declared itself to have international jurisdiction. The chamber based its decision on Section 131(1) and (2) of the Collecting Societies Act (VGG), which provides for a special venue for collecting societies. Accordingly, such disputes can be consolidated in Munich, even if parts of the infringing acts took place abroad.
Because the training took place in the U.S., the court was even required to examine U.S. copyright law under the country-of-protection principle—specifically, the fair use doctrine. The result: not applicable. Since the works were recognizable in the output, the fair use exception did not apply. Even under U.S. law, Suno would have had to obtain a license. For GEMA CEO Dr. Tobias Holzmüller, this is a key point: “If systems are operated in Europe, legal action can also be brought before European courts.”
Second Victory for GEMA Against an AI Provider
The Suno ruling is not GEMA’s first success in the realm of AI. Back in November 2025, the Munich Regional Court had already ruled in favor of the collecting society against OpenAI—that case involved the reproduction of protected song lyrics by ChatGPT. OpenAI has appealed this ruling to the Munich Higher Regional Court. With the Suno case, GEMA is now expanding its focus from song lyrics to the composition itself—that is, its core area of expertise.
If you want to dive deeper into the world of music AI, you’ll also find an overview of Google’s AI music generator, Lyria 3 Pro, on our site.
FAQ: GEMA vs. Suno
Who won the GEMA vs. Suno lawsuit?
GEMA. On July 31, 2026, the Munich I Regional Court largely ruled in favor of the collecting society (Case No. 42 O 763/25).
What exactly was Suno prohibited from doing?
Suno is prohibited from reproducing six protected works without a license or using them to train its AI. In addition, the company must disclose information about revenues and pay damages.
What does “memorization” mean in this context?
The court assumes that the training works are stored in the AI model in such a way that they can be reproduced. It considers this storage to constitute reproduction under Section 16 of the German Copyright Act (UrhG).
Is the ruling final?
No. The ruling is not yet final; Suno can file an appeal.
Why was the lawsuit filed in Germany even though Suno trains its AI in the U.S.?
The court declared itself internationally competent under Section 131 of the German Code of Civil Procedure (VGG) and examined the legal situation under both German and U.S. law.
Conclusion: A precedent with far-reaching implications
The GEMA-Suno ruling is more than just a victory for a collecting society. It is the first European ruling to establish a licensing requirement for training music AI—and it does so with surprisingly clear reasoning. The fact that a German court affirmed jurisdiction over training conducted in the U.S. and even examined fair use is likely to attract attention far beyond Munich.
My assessment: It will now become difficult for AI providers to hide behind the excuse that “we’re just storing patterns.” If a model can reproduce entire songs, the argument that it’s purely statistical is hard to sustain. The ruling isn’t yet final, and an appeal is all but certain. But the direction is clear: Anyone who makes money off the creativity of others will have to pay for it. The written reasoning behind the ruling will now be of great interest—especially the passages on international jurisdiction and the fair use assessment.
