GEMA v Suno: a significant decision for AI music and copyright

On 31 July 2026, the Munich Regional Court handed down an important decision concerning the use of copyright-protected music in the development of generative AI.

The proceedings were brought by the German collecting society GEMA against Suno, the company behind one of the most prominent AI music-generation platforms.

The Court found in favour of GEMA, holding that Suno had infringed copyright by unauthorised use of protected musical works. The decision marks another significant development in the ongoing debate over whether AI developers can use existing copyright works to train commercial generative AI models without first obtaining permission from the relevant rightsholders.

For the music industry, the decision could have important consequences.

What was the GEMA v Suno case about?

Suno operates a generative AI platform which allows users to generate music from relatively simple text prompts.

As with other generative AI systems, developing this technology requires substantial quantities of existing material from which the model can learn the characteristics and structures necessary to generate new content.

GEMA alleged that Suno had used copyright-protected works from its repertoire without obtaining the necessary licences.

The proceedings concerned a number of well-known songs represented by GEMA, including Forever Young, Mambo No. 5 and Daddy Cool.

GEMA argued that the similarities between those copyright-protected works and material generated by Suno demonstrated that the protected works had been used in connection with the development of Suno’s AI model.

The Court agreed with GEMA and found that Suno had processed the relevant works without the necessary rights.

Further to this, Suno was ordered to cease the relevant uses and disclose information concerning revenues generated through its services. GEMA is also entitled to pursue damages arising from the infringement. The judgment remains capable of appeal, and it will therefore be important to see whether the decision is upheld.

Why is the decision significant?

The importance of the case extends beyond the individual songs relied upon by GEMA.

One of the central legal issues surrounding generative AI is whether the use of copyright-protected material as training data requires permission from the relevant rightsholders.

AI developers require very substantial datasets to train generative models, and in the music industry, those datasets may include compositions and sound recordings which remain copyright-protected.

The works of musicians, songwriters and other creators can contribute to the development of commercial AI products which are subsequently capable of generating new music at significant scale. Those products may then compete within the same market as the creators whose works contributed to their development.

The Suno decision provides further support for the position that the process of AI training does not, by itself, place copyright-protected material outside the ordinary framework of copyright law.

And thus, we gain some clarify on an important principle: the fact that a work is being used to develop an AI model does not necessarily mean that the rights of its owner cease to apply.

What does this mean for AI training?

In spite of the above, the decision does not mean that copyright-protected music cannot be used to train AI models. Instead, its wider significance may lie in the development of a licensing framework for AI training.

The music industry has always relied heavily upon licensing. Businesses wishing to exploit compositions or sound recordings will ordinarily need to identify the relevant rights, establish who controls them and obtain the necessary permissions, and so we expect that use of music as AI training material would be approached in much the same way.

Interestingly, GEMA itself has already taken steps in this direction. Shortly before the Suno judgment, GEMA launched “PLAI by GEMA”, a dataset containing approximately 178,000 sound files across more than 60 genres which is specifically intended to provide appropriately licensed material for certain AI training applications.

That development illustrates an important distinction in the current debate. The issue is not necessarily whether AI developers should be prevented from using music to train their systems. Rather, the question is whether they should be permitted to use copyright-protected music commercially without first obtaining the appropriate rights.

The GEMA decision provides further support for the latter being treated in the same way as other commercial uses of copyright.

What does the decision mean in the UK?

The Munich Regional Court’s decision concerns German and European copyright law and does not determine the position in the UK.

The application of UK copyright law to AI training remains the subject of considerable debate, including the extent to which existing copyright exceptions can apply to the use of protected works in training datasets.

Nevertheless, decisions of this nature are relevant to the wider direction in which the market is developing.

Generative AI platforms operate internationally, as do the music businesses and rightsholders whose material may be incorporated into training datasets. A growing body of litigation establishing that permission is required in certain circumstances will inevitably influence how AI companies approach the acquisition and licensing of training material.

For UK artists, songwriters, labels and publishers, the case also highlights the importance of understanding how existing agreements deal with AI.

In particular, rightsholders should consider who has authority to permit works to be used for AI training and how any income generated from those licences is to be distributed.

Similarly, AI developers and businesses incorporating generative AI into their products should be able to establish the provenance of their training material and the rights upon which they are relying.

Where does this leave the music industry?

The GEMA v Suno decision will not settle the wider debate around AI and copyright. Further litigation is ongoing across several jurisdictions, and the German decision may itself be appealed.

However, it represents another significant step towards answering one of the central questions facing the creative industries: whether the development of generative AI creates an entirely new category of use which sits outside established copyright principles, or whether existing principles of ownership, permission and licensing continue to apply.

The recent decision suggests that, at least in Europe, courts may be increasingly reluctant to accept that AI training should operate outside those established principles. If that approach continues, the practical consequence is unlikely to be the end of AI-generated music.

Instead, it may accelerate the development of a market in which access to high-quality copyright-protected training material is licensed, with creators and rightsholders participating in the value generated from the use of their works.

For the music industry, that distinction could prove extremely important.

If you’d like to discuss this decision, or similar, and how they could impact you – please get in touch with our Creative, Digital and Media team.