Munich Regional Court I largely sides with Gema in dispute with AI music company Suno

The case is part of a broader debate over how copyright and training-based AI systems are to be reconciled. Comparable disputes are also being fought in other jurisdictions, for example in the US and in the UK. Industry associations are observing the development with great interest, as the underlying issues of principle have effects beyond the individual case.

What is the dispute about?

Gema represents the rights of composers, lyricists and music publishers in Germany. Suno operates a service with which users can generate songs via text input that are then playable in full length. At the heart of the dispute is the question of whether training such a model with protected music constitutes a copyright-relevant act of reproduction and whether the generated results can be classified as so-called adaptations or free creations.

According to the account of the parties involved, the court tied its assessment to the technical characteristic that certain features of the originals reappear recognizably in the model's outputs. It took this as an indication that the model has stored not only abstract patterns, but concrete, protected elements from the training data. In doing so, the chamber drew on the principles of German copyright law regarding reproduction and adaptation.

In earlier statements of its own, Suno had argued that training was based on publicly accessible data and that the generated results were independent creations. The court did not follow this line on the key points and accepted Gema's argument.

Reasoning of the court

The ruling is not yet final. Suno can appeal the decision. Until a final ruling by higher courts, the exact scope of the court's reasoning remains the subject of legal debate.

No details about the amount of any potential damages claims or the concrete consequences for Suno's business model in Germany were made public as part of the announcement of the ruling. Observers expect that the question of a licensing obligation for training data will remain on the agenda.

Parallel to the proceedings in Munich, discussions are underway on possible licensing models between AI providers and rights holders, as are already being tested in individual markets. At the European level as well, work is being done on guidelines for dealing with generative AI that would be directly relevant to the music industry.

For the German music industry, according to the assessment of industry representatives, the decision is a signal that courts are willing to engage in detail with the technical functioning of trained models. This adds contour to the legal assessment of training data, without there yet being a definitive ruling by the highest court.

Look at further proceedings

The ruling is part of a series of decisions dealing with AI and copyright. In the US, comparable lawsuits have been filed against other providers; in the UK, investigations into AI companies' training practices are ongoing. In Japan as well, questions of how to deal with training data are the subject of political deliberations.

Overall, the Munich proceedings show that courts in Europe are willing to incorporate the technical side of trained AI models into the copyright examination. A definitive ruling by the highest court on the underlying questions of principle is, however, still pending.

Other major AI providers such as OpenAI also face questions about the handling of training data in several countries. The industry is therefore watching the ongoing proceedings closely, as they may provide guidance for future business models.