A landmark European case in AI music copyright law is set to reach a conclusion on July 31, as the Munich Regional Court is expected to deliver its verdict in a lawsuit brought by Germany’s music collection society GEMA against US-based company Suno AI, which generates songs from text prompts.
In the case filed in January 2025, GEMA argues that Suno trained its AI models on copyrighted music without obtaining a license or compensating its rights holders.
The lawsuit is being closely watched, as it is one of the first major cases anywhere in the world testing how traditional copyright law applies to AI music training and AI-generated musical output.
The proceedings could help determine whether AI music companies must obtain licenses before training their models on copyrighted songs and whether AI-generated music that is too similar to existing songs constitutes copyright infringement.
“Our goal is not to wipe Suno off the face of the earth, but to clearly engage in licensing negotiations at eye-to-eye level, which was not possible with Suno until now,” explains Tobias Holzmüller, CEO of GEMA, which describes the company as “the largest and most commercially successful of all AI music models to date.” He further added, “Suno adopts a much more aggressive trading strategy than others.”
The outcome of the lawsuit is also tied to another central question: How should creators be compensated when their music is used to train generative AI systems?
At the forefront of the fierce battle to protect artists’ rights
Last year, GEMA won a related case against OpenAI in the same court. It found that OpenAI illegally trained its AI models on copyrighted song lyrics and reproduced them through ChatGPT without authorization.
OpenAI was ordered to stop reproducing the functions in the model and to disclose the information and pay damages. But since the company appeals, the decision is not final.
In contrast to that decision, the current lawsuit focuses not on song lyrics, but on entire musical pieces.
One aspect that makes this lawsuit significant is that it was not initiated by a private company or record label, but by a music collecting society, says Martin Senftleben, professor of intellectual property law at the Amsterdam Law School. “Initiatives like this are particularly important because, by definition, you can say they are more inclusive,” Senftleben tells DW. “You can become a GEMA member even if you’re not a very famous high-profile artist or musician. And that means there’s a chance for AI users to bring remuneration to a wider group of rights holders.”
GEMA represents more than 95,000 composers, songwriters and music publishers in Germany as well as more than 2 million rights holders worldwide.
How did GEMA present its case?
GEMA claims Suno trained its AI on copyrighted music without permission or payment to songwriters, composers and publishers represented by the German Music Collection Society. GEMA also says the US-based company is profiting commercially from those works without compensating the artists.
Suno’s music creation platform creates songs through simple text prompts. Users can describe the mood, genre or style they desire and the system synthesizes a structured track – complete with lyrics, vocals and instruments – in less than a minute.
Songs produced by AI systems may be similar to copyrighted songs.
During the court proceedings, GEMA demonstrated this by prompting Listen AI to create tracks with original lyrics of various popular songs in its repertoire, including “Forever Young” by Alphaville, “Mambo No. 5” by Lou Bega, “Daddy Cool” by Boney M, “Cherry Cherry Lady” by Modern Talking and “Etemlos durch die Nacht” written by Christina Bach and popularized by Helen Fischer.
Along with the lyrics, the plaintiff indicated the desired musical style and the title of the work, without specifying any melody, rhythm, harmony or arrangement.
Yet the resulting AI songs are very similar For original works.
Suno acknowledges that the original songs at the center of GEMA’s lawsuit are among the myriad pieces that were used to train the AI model.
AI companies like Suno claim that ingestion of data for AI training can be done based on the “fair use” rule in the US and the “text and data mining exception” rule in the EU.
But there remains huge controversy globally over whether such license-free AI training is permitted under copyright law. In various lawsuits around the world, courts are now examining whether such uses harm the economic market of original creators through unfair competition.
“The novelty of the lawsuit is that GEMA does not focus on the AI training phase,” explains Martin Senftleben. By basing their claims on tests that demonstrate that these AI systems produce songs that have similarities with copyright-protected works from their repertoire, “they bridge this gap between the relevant use now taking place in AI training and the harm they cause to the market at the AI output level,” he added.
Why shouldn’t AI be compared to the human brain?
Anyone who wants to become a good musician has to listen to and learn from existing music.
Similarly, the AI industry claims that this analogy applies to how their models are trained – by absorbing countless creative actions to generate new ones.
But from the perspective of rights holders, AI models are copying their works and exploiting them commercially.
The argument that an AI learns from music and is inspired by it like a human is “completely ridiculous”, says Tobias Holzmüller. “These are concepts that come from the human brain and are designed for the human brain, not for a machine that is capable of digesting millions of sound files in a split second. I mean, just the sheer amount of data – they say it’s data; for us, it’s music – that these machines have swallowed and then digested and are now spewing out, it’s completely incomparable to what’s happening in the human brain.”
Holzmüller summarizes, “It is not logical to humanize these machines using these concepts – and it is certainly not ethical.”
On Suno’s platform globally, users collectively create approximately 7 million songs per day. about 75,000 of them “It’s definitely to the point where it’s no longer sustainable, but really toxic for the music ecosystem,” says Holzmuller.
“If you want to create a sustainable model where you accept that AI music is here to stay, but at the same time, it should not drive human-made music out of the market, then the only way to deal with it is to put a price tag on the output,” says GEMA’s CEO. “Because otherwise you’ll have free AI sounds that will outperform human music in many circumstances.”
For example, music in movie scores or social media has allowed various songwriters to earn a living until recently. For Holzmüller, if licensing increases the cost of AI-generated sounds to such an extent that they become as expensive as human-generated songs, this could provide a level playing field: “Then – this is my firm belief – people will continue to use human music, because it has this extra element.”
Edited by: Brenda Haas