GEMA wins a Suno ruling. The court says it is not final.

GEMA won a substantial first-instance ruling against Suno on July 31, 2026. Munich Regional Court I largely granted its claims involving six musical works. The court explicitly said the judgment was not final. The court's announcement is the starting point, rather than either party's description of victory.

The result supports composers seeking control over uses of their work. Its scope also matters. This case does not establish that every AI song infringes, or that every later Suno model has the same legal position.

What did GEMA win in the Suno ruling?

The judgment concerns six identified works, including “Forever Young,” “Daddy Cool,” and “Atemlos durch die Nacht.” The disputed material concerned the music. Lyric infringement was not the subject of this proceeding.

The full judgment identifies Suno versions 3.5 and 4 in the operative restrictions. It addresses reproduction during training, reproduction within the model, and public communication involving the disputed works.

That specificity protects the reader from an easy mistake. A claim about one model and six works is not a technical audit of an entire service. Later versions, different inputs, and different outputs need their own evidence.

GEMA and Suno: the route to the Munich judgment

Date	Event	Evidence status
January 21, 2025	GEMA says it filed the lawsuit.	The plaintiff's account of filing.
March 9, 2026	The court held a hearing and published both sides' positions.	An official procedural account.
July 31, 2026	The court largely ruled for GEMA.	A first-instance judgment, described as not final.
October 7, 2026	DeadMod checked the linked primary sources.	A reporting date, not a new court event.

GEMA's statement presents the result as a victory for licensing and creator payment. GEMA is the successful claimant at this stage. Its broader claims about international significance remain its interpretation.

The court's earlier hearing account records Suno's objections. Suno disputed recognizable copying and argued that its model learned patterns. It also raised fair use, jurisdiction, and German text-and-data-mining defenses.

Why the outputs mattered

The court described prompts containing lyrics, titles, and style requests. It said the prompts did not specify melody, harmony, rhythm, or arrangement. It concluded that recognizable musical expression appeared in outputs from the models at issue.

The hearing account also records repeated prompting. Its counts differ by work, from four prompts for one example to 176 for another. That detail matters because a selected exhibit does not tell a reader how often ordinary users encounter the same result.

Our analysis: reproducible examples are more useful than an unsupported claim that a model sounds familiar. The prompt, model version, generated file, and comparison method all affect what an example can establish. A strong creator-rights argument benefits from those details.

A listener's reaction can identify a concern. It cannot, alone, establish the training history or the legal cause of the similarity. This case involved evidence and judicial findings concerning named works.

Impact on users

For musicians making lyric videos, a familiar generated melody deserves attention before it becomes the center of a release. Adding your own visuals does not explain where that musical expression came from. Neither does changing the filename or adding a new artist credit.

Our analysis: preserve the original generation and your later edits as separate files. Keep the model version and prompt history beside them. If you need a rights assessment, that record gives your adviser something concrete to review.

The ruling concerns Suno's responsibility in the circumstances before the court. It is not a blanket immunity for users. It also does not require the conclusion that every subscriber committed an infringement.

Artists who own the source song can still make their control visible. Keep writing drafts, session files, performer permissions, and artwork licenses together. These records serve a practical purpose when distributors, collaborators, or video clients ask about a release.

German jurisdiction is not a US precedent

The court applied German law to relevant German acts and assessed US law for training acts in the United States. Its announcement explains the jurisdictional basis it accepted for GEMA's claims.

A German court's assessment of US fair use is not a binding decision from a US federal court. It is therefore misleading to call this a final American ruling against AI music training. The location of the court and the law it applies are separate facts.

Our analysis: cross-border disputes make broad “legal everywhere” marketing especially unhelpful. A creator distributing internationally needs clear permissions for the actual release. A favorable result in one territory does not silently clear uses in all others.

What remains open after July 31

The official announcement says the judgment is not final. This article does not establish the current appeal docket, a final damages amount, or the outcome of any later challenge. No such outcome is inferred from the passage of time.

It also does not treat the decision as a ruling on v6. The older versions named in the judgment are a material limit. Future reporting needs to identify the specific model, work, claim, and procedural stage.

The broader creator question remains practical: who agreed to the use, who receives payment, and what control survives the agreement? Those questions remain useful even after litigation ends.

Our Suno lyrics report explains why an automated rejection is a different event from a legal judgment. Our custom-model report addresses the separate problem of keeping projects usable through model changes.