The Defense a Court Just Rejected
Suno's defense in Munich rested on a single idea: if the EU AI Act already requires a public summary of training content, a separate copyright license should not be necessary on top of it. The Regional Court of Munich I rejected that argument entirely on July 31, 2026, ruling that "the EU AI Act does not alter the applicable EU copyright framework" and that its Article 53 transparency duties "complement the requirements under the DSM Directive and the InfoSoc Directive" rather than replace them. Regulatory disclosure and copyright clearance are two separate obligations, and passing one does not excuse a company from the other.
This is the argument that should worry any business buying a generative AI tool, not just Suno. A vendor's AI Act compliance page, its published training-data summary, its transparency certification, none of it says anything about whether the underlying content was licensed. The two questions are unrelated, and this ruling is the first to say so on the record.
What Munich Actually Found
The case, filed by the German collecting society GEMA over six musical compositions, produced findings against four separate acts: reproduction for training purposes in the United States, reproduction through memorization inside the model itself in Germany, communication to the public through offering the model, and reproduction and communication to the public through the outputs it generates. The court ordered Suno to disclose the scale of its use of GEMA's catalogue and to pay damages, with the exact amount still to be determined.
The most consequential finding sits underneath the headline: German copyright law reached conduct that happened entirely on US soil. The court based its jurisdiction on a venue rule specific to collecting societies, then applied US copyright law to the training acts themselves and rejected Suno's fair use defense, distinguishing the case from prior US rulings where the training material was not recognizable in the model's outputs.
The Bar Keeps Getting Lower
This is GEMA's second win against a generative AI provider in under a year. In November 2025, the same Munich court ruled against OpenAI over its ability to reproduce song lyrics on request, establishing a rule that carries forward: a claimant does not need "direct technical evidence demonstrating that a specific work can be concretely identified in the parameters of the model." Suno now adds a second precedent on top of that one, extending German liability to training conducted abroad.
Two rulings against two different companies, from the same court, in under twelve months, is not a coincidence worth ignoring. Every AI vendor selling a generative tool into Germany, whatever it generates, is now operating against a track record of German courts finding copyright liability with a lower evidentiary bar and a longer jurisdictional reach than most vendors assumed a year ago.
The Question to Ask Your AI Vendor
Before this ruling, a procurement team could reasonably treat a vendor's AI Act transparency page as a proxy for its overall compliance posture. That shortcut no longer holds. The two questions, has this vendor met its AI Act disclosure duties, and has this vendor cleared copyright on the content it trained on, now need separate answers, and a vendor's confident yes to the first is not evidence of anything about the second.
For a European business licensing a generative AI tool for marketing copy, product images, code, or music, the practical move is to ask directly whether training content was licensed or falls under a specific exception, rather than accepting an AI Act compliance certificate as an answer to a question it was never designed to answer.
Servola Journal
We do this for everyone trying to keep up with what technology is doing to our lives. The people who build it, and the people it happens to. The Servola Journal exists so that what we learn belongs to all of them.
Nobody pays us for this. No ads, no paywall, free to everyone. We just believe that understanding what's happening to all of us shouldn't depend on who can afford to pay for it.
If it gave you something today, tell us to keep going. Follow us, leave a like, or write a positive comment. We read every one, and they are what keeps us going.
Read next: OpenAI Wants Its Own Incident Clock. Brussels Started One First. | Your AI Vendor Just Got a Letter From Brussels



