On 31 July 2026 the 42nd Civil Chamber of the Regional Court Munich I, the chamber that specialises in copyright, issued its judgment in the case brought by the German collecting society GEMA against the AI music generator Suno, case number 42 O 763/25. The court granted GEMA’s claims for injunctive relief, disclosure and accounting, and damages for the most part.
The reasoning covers both ends of the pipeline. On the input side, the court found a violation of the reproduction right under Section 16 of the German Copyright Act (UrhG), on the basis that training data was memorised within the models held on German servers, and it rejected the argument that the text and data mining exception in Section 44b UrhG covered that reproduction. On the output side, the court held that the generator’s outputs contained recognisable original elements, so operating the music generator itself infringed the right of public performance under Section 15(2) UrhG. For the training that took place in the United States, the court applied US law and concluded that fair use did not apply, distinguishing the case from earlier decisions where training data was not substantially reproduced in the outputs.
The judgment is not final - the court’s own notice records that it is not yet legally binding and can be appealed - so the holding is a first-instance result rather than settled European law.
For a business leader, the significance is that a European court has now separated the two questions that AI copyright disputes usually blur together. Memorisation inside model weights was treated as a reproduction in its own right, and the text and data mining exception did not rescue it. Any organisation licensing or reselling generative output built on scraped creative works should expect diligence questions about what the model memorised, not only about what data was collected.