AI & Music

A Munich Court Rules Against Suno in GEMA Case: What It Decided

By the Starchild Music team · October 19, 2026 · 5 min read

A stately courtroom bench in warm wood with an old brass scale on it, a ribbon of audio waveform draped over one pan, and a small paper model of a Bavarian church beside it

The short answer

On July 31, 2026, the Munich Regional Court ruled that Suno infringed the rights of GEMA members by training on and reproducing their songs, a first for a European court on training done in the United States. Suno said it disagrees and is considering an appeal, and the judgment is not yet enforceable. The sources reviewed report no appeal filed yet, so the legal picture can still change.

Courts have started to rule on AI and music. On July 31, 2026, the Munich Regional Court ruled for GEMA, the German collecting society, in its case against Suno. It is the second GEMA ruling against an AI company, after a November 2025 decision involving OpenAI.

The ruling gets less attention in the United States than the American cases, but it addresses a question that matters everywhere: can a company in one country be held responsible in another for how it trained a model?

What the court decided

GEMA says the court found Suno trained its AI systems on works from GEMA's repertoire in the United States, then stored and reproduced them in Europe, without licenses. The 42nd Chamber found this breached both U.S. and German copyright law.

The works at issue included "Forever Young," "Atemlos durch die Nacht," "Mambo No. 5," "Rasputin," "Big in Japan" and "Daddy Cool," which Music Week lists as six songs.

The JUVE Patent report describes the orders. The court told Suno to stop the unauthorised reproduction of the works and to stop using them to train the model, and to disclose revenue information. The amount of damages is to be decided later.

The reasoning

JUVE Patent says the decisive point was memorization: the court found the similarities between Suno's outputs and the original songs were "too extensive and specific to be coincidental." GEMA's statement says the system's output matched the originals "in terms of melody, harmony and rhythm."

On American law, the court applied U.S. copyright rules to the training that took place in the United States and distinguished earlier American decisions in the Bartz and Kadrey cases. JUVE reports the court noted that simple prompts produced outputs substantially similar to the original works.

It also records a finding that Suno used stream-ripping to extract works from YouTube, getting around a technical protection measure. These are the court's findings in a first-instance ruling, not final law.

The earlier OpenAI decision

The Suno ruling follows GEMA's case against OpenAI. On November 11, 2025, the same court ruled that OpenAI had used lyrics from nine popular German songs to train its language models without a license, according to coverage compiled by Media Laws. The court held that the lyrics were "reproducibly fixed in the models" through memorization, and that the text and data mining exception did not apply because "the memorization of the lyrics exceeds this generation of information."

It ordered OpenAI to stop reproducing the lyrics in GPT-4 and GPT-4o and awarded damages, and it found OpenAI had acted negligently. Together the two decisions rest on the same idea: when a model stores a work closely enough to reproduce it, the court treated that as copying.

What Suno has said

Music Week reports that Suno said it "disagree[s] with today's ruling" and is "evaluating all available options, including an appeal." JUVE notes the judgment is not yet enforceable.

The sources reviewed do not report that an appeal has been filed. For comparison, GEMA says OpenAI has appealed its November 2025 loss to the Higher Regional Court in Munich.

Why the location of training matters

Music Business Worldwide, in an August 4 analysis of the EU AI Act, quotes the principle that unauthorized AI training "can infringe copyright even where the training itself takes place outside the EU." It calls the Munich decision the first major European ruling to extend liability that way, grounded in German collecting-society law rather than the AI Act itself.

MBW also explains that rights holders can reserve works from text and data mining by machine-readable means under Article 4(3) of the EU's 2019 Copyright Directive. Where an opt-out exists, a model cannot lawfully be trained on those tracks for the EU market without a license. Sony Music and Warner Music Group have already exercised that right, according to MBW.

The regulatory backdrop

The ruling lands while the EU AI Act's obligations are phasing in. MBW's August 4 analysis says the content-labeling rules became enforceable on August 2, 2026, that providers must publish detailed summaries of training data, and that existing systems have until December 2, 2026 to add machine-readable marking to AI output.

MBW adds that penalties reach up to 15 million euros or 3% of worldwide annual turnover for labeling breaches, enforced by national regulators, with the EU's AI Office enforcing the copyright obligations. The Munich ruling and the AI Act work on different legal footing, but both push toward more disclosure about what models were trained on.

What it means for independent songwriters

This is a first-instance decision, and the sources reviewed do not report a final outcome. Damages remain to be set, and an appeal could change the reasoning. Readers should treat it as an important signal rather than settled law.

The ruling directly concerns works in GEMA's repertoire, so it does not by itself give every songwriter a claim. Collecting societies are the parties that bring such cases for their members, which is one reason membership and registrations matter.

It does suggest that a court outside the United States may treat model memorization as copying even when American fair use arguments are raised. That sits alongside the U.S. appeals-court decision we covered in the Third Circuit ruling on AI training, and next to the licensing route Suno has taken, described in Suno v6 and the new lawsuit. Songwriters should treat each decision as one data point, not a settled rule.

For artists who want songs with documented authorship, the Starchild catalog is built on songs by professional songwriters, and each license certificate names the writers with their PRO and IPI details. Read how licensing works for the steps.

Sources