MUNICH – A German court has ruled that AI music company Suno violated copyright rules by using songs represented by the collecting society GEMA without a license, adding a significant European decision to the global dispute over training generative models on protected work.

The Munich Regional Court found that Suno lacked the rights to process the songs and ordered the company to disclose revenue connected to the disputed use, Reuters reported. The case was heard under file number 42 O 763/25.

GEMA, which represents more than 100,000 composers, lyricists and music publishers in Germany, filed the lawsuit in January 2025. The organization argued that Suno trained its systems on protected songs without compensation and could reproduce outputs closely resembling those works.

The six songs identified by GEMA included “Forever Young,” “Atemlos,” “Mambo No. 5,” “Rasputin,” “Big in Japan” and “Daddy Cool.” In its statement on the decision, GEMA said it had documented similarities in melody, harmony and rhythm between generated outputs and the original music.

The court also addressed the international structure of AI development. GEMA said the ruling found that training conducted in the United States could still violate U.S. copyright law and that operation of the resulting system in Europe could support a case in a European court.

Suno had disputed that it owed licensing payments for the training activity, according to GEMA. The company is also facing litigation from major record labels in the United States, where courts are still weighing how existing copyright doctrines apply to generative models.

The Munich decision does not settle those U.S. cases, and the legal rules governing AI training remain contested across jurisdictions. It does, however, strengthen the argument that commercial music-generation systems should negotiate licenses before using protected catalogs.

That distinction is becoming central to the next phase of AI music. Record labels, publishers and technology companies have begun developing licensed services and revenue-sharing agreements, while unlicensed training remains the subject of lawsuits. The business question is shifting from whether generative music will exist to which systems can document permission, compensation and control.