Analysis
A Munich Regional Court ruled this week that Suno is liable for copyright infringement, finding the AI music company's model memorized and reproduced protected songs from German rights organization GEMA's catalog -- the first major European verdict to attach real financial liability to AI training data rather than settling the question on fair-use grounds alone.
The court found Suno breached both US and German copyright law: by training on GEMA-repertoire songs sourced in the US, and then by storing and reproducing them within Europe. Evidence presented at trial showed Suno's model had memorized and could reproduce six GEMA-represented tracks verbatim, including "Forever Young" and "Daddy Cool," out of a training set built from more than two million scraped songs.
Critically, the court found liability on two separate legal grounds: storing the protected songs inside the model itself infringes the reproduction right, while serving outputs built on them to users separately infringes the right of making works available to the public. That's a materially narrower and harder-to-dodge theory than the broad "training on copyrighted data is inherently infringing" arguments still working through parallel cases in US courts -- it's a finding built on proof that specific, named songs came back out of the model closely enough to count as reproduction.
Suno must now disclose its revenues to GEMA and pay damages in an amount still to be determined, and has said it disagrees with the ruling and is evaluating its options, including an appeal. The decision lands as the first concrete signal that European courts are willing to impose real financial consequences on AI training practices rather than waiting for legislative clarity.
What to watch: whether Suno appeals and on what grounds, how the damages calculation plays out once revenues are disclosed, and whether other European rights organizations bring similar suits against AI music or video generators now that a court has established a working legal theory.
