A German court has convicted Suno AI of violating music copyrights.

GEMA, the collective management society for music rights in Germany, representing more than 95,000 composers, authors and publishers in the country as well as more than 2 million rights holders worldwide, had taken legal action in January 2025. Its central grievance: Suno had allegedly used protected works to train its algorithms without obtaining licenses or compensating the creators concerned.

The court ruled in favor of GEMA. Suno will have to disclose the illegally generated revenue and pay damages, the precise amount of which will be determined later. The American company contested the decision and announced that it would explore all available legal avenues, including an appeal. During the proceedings, it had argued that the German court lacked jurisdiction to rule on model training activities conducted exclusively within the United States.

“This is a verdict of global significance,” said Tobias Holzmüller, CEO of GEMA, clarifying that the action’s objective was not to eliminate Suno from the market, but to open licensing negotiations “on an equal footing, which had not been possible until now” with the company. He described Suno as “by far the largest and most commercially profitable AI music model,” and criticized its strategy as being far more aggressive than that of its competitors.

To support its complaint, GEMA submitted text queries to the platform containing the original lyrics of several popular songs from its repertoire: "Forever Young" by Alphaville, "Mambo No. 5" by Lou Bega, "Daddy Cool" by Boney M., "Cheri Cheri Lady" by Modern Talking, and "Atemlos durch die Nacht," written by Kristina Bach and popularized by Helene Fischer. Without providing any melody, rhythm, or arrangement details, the queries produced tracks deemed very close to the originals. Suno also acknowledges that these songs are among the many works used to train its model.

The unique legal aspect of this case lies in the strategy adopted by GEMA: rather than attacking the training phase of the model, the organization focused its argument on the outputs generated by the AI, demonstrating their resemblance to copyrighted works. Martin Senftleben, professor of intellectual property law at Amsterdam Law School, emphasizes that this approach "bridges the gap between the relevant use during training and the commercial harm caused at the level of the AI's outputs."

This case is part of a global legal debate on the legality of training AI models using copyrighted works. Companies in the sector invoke the "fair use" principle in the United States and the text and data mining exception in Europe to justify ingesting unlicensed content. This interpretation remains hotly contested in several courts around the world.

GEMA had already won a case last year before the same court in a separate matter against OpenAI, which was found guilty of training its models on copyrighted song lyrics and reproducing them via ChatGPT without authorization. That decision is under appeal and therefore not final. The proceedings against Suno, however, concern complete musical pieces, not just lyrics.

Suno generates approximately 7 million tracks per day on its platform, with some 75,000 of these ending up on streaming services daily. For Tobias Holzmüller, this volume makes the current model "toxic for the music ecosystem." He argues that AI productions should be subject to a licensing fee equivalent to that of human-generated music, to prevent free machine-generated content from driving creators out of the market. "If you want a sustainable model where musical AI coexists without crushing human music, the only solution is to put a price on what it produces," he summarizes.