Munich Court Halts AI Music Generator Suno Over Copyright Infringement
Munich, August 4, 2026 – The Munich Regional Court has issued a significant ruling against the AI music generator Suno, prohibiting its unauthorized use of copyrighted songs for training and reproduction. The 42nd Chamber found clear violations of copyright law, ordering Suno to cease its activities, provide information, and pay damages. A key aspect of the decision was whether the AI models store music data rather than merely “learning” from it. The judgment is not yet legally binding, and Suno is reportedly reviewing its options.
The court’s decision sends a strong signal regarding the use of copyrighted music in AI workflows. The 42nd Chamber prohibited Suno from using protected songs without authorization, citing clear copyright infringements identified during the proceedings. According to the court’s reasoning, it is not just about theoretical training data, but the observable effect that the model was able to recognizably reproduce specific songs after training. Presiding Judge Elke Schwager sided with Suno only on minor points, while the core claim of the rights holders was upheld.
The Crucial Question: Learning vs. Storing
Central to the case was the question of whether songs are merely statistically processed during training or actually stored. For copyright law, this distinction is critical: if protected content is stored or duplicated in a reproducible form, a copyright-relevant reproduction quickly occurs. Suno had previously argued that its models “only learn” but do not store. The court did not accept this view, stating that it could not be that songs are initially used for training and then recognizably reproduced without copyright losing its significance in between.
The specific impetus for the proceedings involved six musical pieces or parts thereof, which the court cited as well-known references: “Atemlos,” “Daddy Cool,” “Mambo No. 5,” “Big in Japan,” “Forever Young,” and “Rasputin.” In tests conducted by GEMA, the court was convinced that the model was capable of reproducing these titles recognizably during generation. The complexity of the proceedings was particularly noteworthy: the oral justification of the judgment alone lasted an hour and a half, as several central legal and technical issues converged – from the storage discussion to jurisdiction and accountability.
Jurisdiction and Fair Use: A Detailed Legal Examination
The legal framework of the proceedings was also thoroughly examined. The jurisdiction of the Munich court was disputed because the AI training took place in the USA. However, Judge Schwager expressly affirmed jurisdiction, citing a special rule allowing collecting societies to sue at the location of an alleged infringement and to include other infringements from other locations. The examination of US law was equally clear: the principle of “fair use” did not apply here, partly because the music data was accessed despite download protection and used for commercial purposes. The fact that it involved commercial use and not purely non-commercial research typically plays a central role in legal assessment.
The court also formulated a noticeably strict attribution of the generated results. On the contentious point of who is responsible for the AI output, the chamber made a clear statement: since Suno is responsible for all preceding steps, the provider also bears responsibility for what its music generator produces. Thus, the judgment’s logic in the overall chain favors the provider rather than the individual user who merely retrieves the output. GEMA CEO Tobias Holzmüller expressed satisfaction, calling it “a good day for authors and copyright holders, but probably also for European culture.” He stated that the court had “very clearly” described Suno’s activities as intellectual property theft.
Broader Implications for AI and Copyright
Looking beyond this specific case, courts increasingly face the challenge of addressing the technical plausibility of “just learning, not storing.” The judgment, according to its own line, is reminiscent of a decision from last November, in which Elke Schwager, also as presiding judge, reached a storage classification in the context of OpenAI and ChatGPT. That case involved protected song lyrics that the chatbot could output identically or largely identically; there, too, the court assumed storage. In addition, there are international legal approaches, for example, from Great Britain concerning Getty Images, where AI models had “learned” but, according to the court, had not stored in a way that conflicted with copyright – a contrast that shows how much it depends on technical evidence and specific functionalities.
For practical purposes, the negotiating position between rights holders and AI providers is changing. GEMA aims to compel Suno and other companies to acquire licenses to ensure the legal use of protected content in the future. Suno views the outcome of the proceedings differently, stating that it “disagrees with today’s ruling,” which it believes is based on a “misrepresentation” of how the technology works and is used, as well as how US law applies; the company is exploring “all available options,” including an appeal. Peter Maffay, who is involved with GEMA and was present at the verdict, emphasized the artists’ perspective: the judgment protects their claims and rights and is not against AI, but against the lack of participation of creatives in the results derived from their substance. Culture Minister Wolfram Weimer added that innovation and creativity can only be meaningfully linked if a legal framework adequately compensates rights holders and enables them to participate in value creation through AI. Since the judgment is not yet legally binding, the legal dynamics remain open; technically and economically, however, the direction is clear: AI providers will increasingly have to demonstrably justify how training data is processed, and not just how they imply it.
Source: IT BOLTWISE®
