Suno loses in Munich as AI music training faces license test

A European court has put AI music training on a sharper legal path. On July 31, 2026, GEMA said the Munich Regional Court ruled in its favor against Suno, finding that the U.S. AI audio company infringed German and U.S. copyright law through training on well-known songs from the GEMA repertoire, as well as through storing and reproducing those works in Europe.

The primary account is GEMA's July 31 press release. GEMA filed the case on January 21, 2025, after seeking compensation for the use of its members' works. The collecting society says Suno's generative audio system could produce playable tracks confusingly similar to “Forever Young,” “Atemlos,” “Mambo No. 5,” “Rasputin,” “Big in Japan,” and “Daddy Cool.” It also says it documented outputs that substantially matched the melody, harmony, and rhythm of the original works.

The most important part of the ruling, as described by GEMA, is not only similarity. GEMA says the 42nd chamber of the Munich court held that Suno needed a license for systematic use and commercial exploitation of GEMA repertoire. The court also addressed jurisdiction: according to GEMA, this was the first European court decision on systems trained in the United States, and the Munich court found that U.S. copyright law also required Suno to obtain a GEMA license.

That detail matters for AI developers. Many copyright disputes around model training turn on where training happened, whether models merely analyze works or store recoverable expression, and who is responsible when users generate infringing outputs. GEMA's summary says the court accepted evidence that Suno's systems stored and emitted content closely aligned with protected songs. If that reasoning survives appeal, it could raise the licensing pressure on AI music services offered in Europe even when their training infrastructure sits elsewhere.

GEMA connects the Suno ruling to its November 2025 win against OpenAI in Munich, where the court found that OpenAI stored and reproduced protected song lyrics without licenses. OpenAI has appealed that earlier decision. The Suno case extends the fight into playable audio, which is closer to GEMA's core licensing business and more directly relevant to composers, publishers, and music platforms.

There are still limits to what can be concluded today. The public source is GEMA's statement, not the full judgment. Suno may appeal, and the final damages and operational consequences are not settled in the available materials. Even so, the ruling gives rightsholders a concrete legal argument: AI music products may need to account for licensed training data and output controls as product costs, not optional policy features.

The next checkpoints are the written judgment, Suno's response, any appeal in Germany, and whether other European collecting societies use the same theory. For AI music startups, the practical question is now simple: can they prove their training and generation pipeline respects music rights before courts or licensors force that accounting later?