AI Law - International Review of Artificial Intelligence LawCC BY-NC-SA Commercial Licence ISSN 3035-5451
G. Giappichelli Editore

02/08/2026 - Munich Court Finds AI Music Generator Suno Reproduced and Infringed Popular Songs (Germany)

argument: Notizie/News - Intellectual Property Law

Source: The Decoder

GEMA sued Suno for injunction, disclosure, and damages and prevailed on most counts in a Munich court concerning six well‑known musical compositions; the dispute focused on musical compositions only, not lyrics. The court determined that all six tracks were reproducibly contained in Suno’s version 3.5 and 4 models, characterizing this as memorization where a model stores specific training content that can later appear in outputs.

For its probe, GEMA entered each song’s complete lyrics, musical style, and title into Suno’s generator without specifying melody, harmony, rhythm, or arrangement, and the court found outputs that reproduced identifiable original elements; it rejected Suno’s defense that the models only learned generalized patterns. The court held Suno — which selected the training data and built the architecture — responsible for the infringing outputs rather than the users who typed prompts, described GEMA’s prompts as “simple and open‑ended,” and warned that the ruling could affect other AI music services; the judgment is not final. The court ruled Germany’s text‑and‑data‑mining exception did not cover the memorization it found and, invoking a special rule for collecting societies, asserted jurisdiction over training acts in the United States and applied US law to those acts. Distinguishing the case from the US Bartz and Kadrey proceedings where training outputs were not substantially accessible to users, the Munich court concluded that all factors of the US fair use test (as discussed in Warhol) weighed against Suno because simple inputs produced results “substantially similar” to the originals. According to the court’s press release, Suno allegedly used “stream‑ripping techniques” to extract music from YouTube and bypassed YouTube’s “Rolling Cipher,” a fact the decision flagged as potentially significant; the court also noted that research has shown comparable memorization effects with books. The ruling emphasized unresolved questions about how often a work must appear in training data and how targeted prompts must be before reproduction and liability are proven, and it stressed that the decision is provisional pending appeal.