> ## Content Index > Fetch the complete content index at: https://theseguysknow.io/llms.txt > Use this file to discover other available public pages before exploring further. # Suno Ripped Songs From YouTube. A German Court Says It Has to Pay - URL: https://theseguysknow.io/suno-copyright-case-youtube-songs/ - Published: 2026-08-01T08:03:31.000Z - Updated: 2026-08-01T08:03:31.000Z - Description: Suno says its AI learns music like a child. A German court says it ripped six protected songs from YouTube and kept enough to reproduce them. - Author: Mike Hazard - Tags: Things to Learn, AI News Suno needs existing music to train an AI that can generate convincing songs. That does not make every track on YouTube free training material. The Suno copyright case is about the way the company obtained and used six protected songs. According to the Munich Regional Court, Suno extracted them from YouTube with stream-ripping tools, bypassed the platform’s download protection and retained enough of the compositions inside its models for recognisable musical elements to return in new outputs. The court largely upheld GEMA’s claims for an injunction, financial disclosure and damages. Suno disagrees with the ruling and may appeal, so the case is not finished. Even so, the decision is a serious problem for the company because it deals with specific evidence rather than a vague argument about whether AI is good or bad for musicians. ## Quick Answer The [Munich Regional Court found](https://www.justiz.bayern.de/gerichte-und-behoerden/landgericht/muenchen-1/presse/2026/16.php?ref=theseguysknow.io) that Suno copied six protected compositions from YouTube for training and reproduced them inside its v3.5 and v4 models. GEMA then generated outputs containing recognisable elements of those songs without specifying their melody, harmony, rhythm or arrangement in its prompts. The ruling covers “Atemlos durch die Nacht,” “Rasputin,” “Big in Japan,” “Forever Young,” the chorus of “Mambo No. 5” and “Daddy Cool.” It does not ban Suno or declare all AI music illegal, and lyric infringement was outside the case. If the judgment survives an appeal, Suno must stop the covered uses, provide information needed to calculate the relevant revenue and pay damages. The amount has not been decided. ## What Is Suno? Suno is one of the biggest AI music generators. A user can describe a song, provide lyrics or ask the platform to write them, and receive a complete recording with vocals, melody, instruments, arrangement and production. Someone with no microphone, instruments or production experience can enter a prompt such as “melancholic 1980s synth-pop about leaving home” and get a listenable track within minutes. Producers can also use it to sketch an idea before rebuilding the song elsewhere, which is why dismissing the whole product as a toy misses what makes it attractive. This is also a serious commercial business. Suno [raised more than $400 million in June 2026](https://suno.com/blog/series-d-announcement?ref=theseguysknow.io) at a post-money valuation of $5.4 billion, while the company said in late 2025 that its community had grown to nearly 100 million music makers. ## What the Court Says Suno Did GEMA, the German collecting society representing the relevant writers and publishers, brought the case over six compositions: - “Atemlos durch die Nacht” - “Rasputin” - “Big in Japan” - “Forever Young” - The chorus of “Mambo No. 5” - “Daddy Cool” According to the [court’s official summary](https://www.justiz.bayern.de/gerichte-und-behoerden/landgericht/muenchen-1/presse/2026/16.php?ref=theseguysknow.io), Suno’s training dataset contained all six works. The company used stream-ripping techniques to extract and copy them from YouTube while bypassing Rolling Cipher, a technical measure intended to prevent direct downloads of the platform’s audio and video. That finding matters because Suno has described its training material in much softer terms. The company says it learns from medium- and high-quality music available on the “open internet,” including copyrighted material. In practice, the court says this included working around a download protection system to copy files into a commercial training process. YouTube gives people access to stream authorised uploads; it does not hand every viewer a licence to copy the underlying music into another product. Suno can argue that training should qualify as fair use, but calling the source “open” does not settle the question of how the files were acquired or what the model later reproduced. ## How GEMA Tested the Models GEMA entered the original lyrics, song title and requested musical style for each work. It did not specify the melody, harmony, rhythm or arrangement, and lyric infringement was never part of the claim. Suno argued that GEMA had forced the results through complex prompts and repeatedly narrowed the model’s choices until it produced something similar. The court rejected that account, describing the prompts as simple and open-ended on the musical details that mattered. There is still an important qualification here: these were targeted tests, since GEMA supplied both the title and original lyrics. Nobody accidentally recreated “Forever Young” after asking Suno for a generic synth-pop track. But the protected melody and arrangement were not placed in the prompt. The purpose of the test was to see whether those musical elements remained recoverable from the model, and the court decided that the similarities were too extensive to be explained by chance. The judgment says the six compositions were reproducibly contained in Suno’s v3.5 and v4 models, which were stored on servers in Germany. By comparing the training works with the outputs, the court found that the models had memorised recognisable content rather than learning only broad ideas about genre, instrumentation or song structure. ## Why Suno’s “Human Learning” Defence Failed Suno has publicly compared model training with a child learning to write rock songs by listening to rock music. In its [2024 defence of its training practices](https://suno.com/blog/future-of-music?ref=theseguysknow.io), the company argued that its system learns styles, patterns and forms—the “grammar” of music—before creating original work. Part of that argument is reasonable. Every musician learns by listening, and copyright does not give anyone ownership of an entire genre, a common chord progression or the general idea of a big pop chorus. A rule that treated every trace of influence as infringement would make ordinary music writing impossible. The Munich case turned on something more specific: recognisable parts of six compositions returned in Suno’s outputs after targeted but musically open-ended prompts. Once that happens, the debate is no longer only about whether a model may learn general patterns from copyrighted work. The scale and method also weaken the comparison with a human listener. A child with headphones does not use automated tools to make millions of digital copies, bypass download protection, process those files through commercial infrastructure and sell access to the resulting system. Suno is a company making technical and business decisions, so pretending that its servers are simply “listening” avoids the hard part of the argument. The court also examined Suno’s claim that copying during training in the United States was protected by fair use. It found that all four factors went against Suno on these facts because substantially similar music remained available through the outputs, distinguishing the case from US rulings involving Anthropic and Meta where the training works were not substantially exposed to users. ## Could Suno Train Legally? Suno cannot build a useful music generator without training on music, but it has several alternatives to copying protected songs without permission. Licensed catalogues, opt-in music, commissioned recordings, properly cleared datasets and verified public-domain material can all provide training data, although assembling enough of it would cost more and take longer. The company has already shown that licensing is possible. In November 2025, Suno [announced a partnership with Warner Music Group](https://suno.com/blog/wmg-partnership?ref=theseguysknow.io) to develop a new generation of models using licensed music, with participating artists able to opt in to certain uses of their names, voices, likenesses and compositions. Suno said in June 2026 that its first model developed with the music industry would begin rolling out in the coming months. Whatever one thinks of the terms Warner negotiated, the agreement ends the claim that a licensed AI music model is technically unrealistic. The real inconvenience is financial. Training first on music gathered from the internet allowed Suno to improve the product, attract a huge user base and reach a $5.4 billion valuation before it had secured permission from every rights holder involved. Licensing earlier would have made that route slower and more expensive, but those are normal business costs rather than a technical impossibility. ## Why Musicians Have a Fair Complaint Everyone else involved in Suno’s business is paid in the normal way: cloud providers charge for computing power, employees earn salaries, investors receive equity and customers buy subscriptions. The musicians whose work helped build the product are somehow expected to accept that their contribution was free because the songs could be found online. That arrangement is difficult to defend, especially in an industry where working musicians already fight over tiny streaming payments. Suno can generate an effectively unlimited supply of new tracks from a system trained on music that took composers, performers and producers years to create. Licensing alone will not guarantee a fair outcome for artists. Major labels may collect large fees while passing little to individual musicians, independent creators will have less bargaining power, and expensive catalogue deals could leave the strongest AI models controlled by a few large companies. Those are arguments for transparent licences and better distribution, though, rather than a reason to pay nothing for the training material. There is no contradiction in finding Suno useful while believing it should compensate rights holders. Producers may use it for sketches, ordinary people may enjoy creating songs with it, and the company may eventually build licensed models that artists genuinely want to join. None of that erases how the earlier models were built. ## What the Ruling Does and Does Not Mean Germany has not banned Suno, and the court did not decide that every AI-generated song infringes copyright. Its judgment concerns six specific compositions, two model versions and evidence that covered the training copies, memorisation inside models stored in Germany and substantially similar outputs. The ruling also does not establish that every use of copyrighted music for AI training is unlawful. A different model trained under another licence, using another dataset or producing outputs with no recognisable source material could present a very different legal case. Suno has said that it disagrees and is evaluating its options, including an appeal. The judgment is therefore not final, and a higher court could revisit the German court’s jurisdiction, its treatment of model memorisation or its analysis of US fair use. If the decision stands, however, Suno will have to disclose financial information connected to the unlawful uses and pay damages that have yet to be quantified. The ruling would also give GEMA and other rights holders stronger leverage when negotiating licences for the catalogues AI music companies need. ## What Happens Next? Suno’s next generation of licensed models may eventually replace the versions examined in this case. That move would deal with future training, although it would not answer what should happen to the value created by older models or the music already used to build them. AI music will continue growing because the product is useful, cheap and increasingly convincing. The argument is now about the terms under which that growth happens, and Suno’s position became much harder once a court found that protected songs were copied, memorised and recoverable from its models. Suno needed real music to make its generator good enough for nearly 100 million people to use. It chose to copy first and negotiate later; the Munich court has now ruled that this decision infringed copyright. ## Common Questions ### Did Germany ban Suno? No. The Munich court ruled on Suno’s use and reproduction of six protected compositions, while the service remains available and the judgment can still be appealed. ### Which Suno models were involved? The court found that the six compositions were reproducibly contained in Suno’s v3.5 and v4 models stored on servers in Germany. ### Did Suno copy the lyrics? Lyric infringement was outside this case. GEMA supplied the original lyrics in its prompts, while the dispute concerned recognisable melody, harmony, rhythm and other musical elements generated around them. ### How can Suno train legally? Suno can use licensed catalogues, opt-in music, commissioned recordings, cleared datasets and properly verified public-domain material. Its partnership with Warner Music Group already provides one route towards licensed models. ### Does this make all AI music training illegal? No. The judgment concerns specific evidence involving six works, YouTube stream-ripping, memorisation and substantially similar outputs, so another dataset or model could produce a different result. ### Will Suno appeal? Suno told [Reuters](https://www.reuters.com/world/german-court-rules-ai-music-firm-suno-broke-copyright-rules-2026-07-31/?ref=theseguysknow.io) that it disagrees with the ruling and is considering all available options, including an appeal. #### Sources - [Munich Regional Court: Judgment summary in GEMA v. Suno](https://www.justiz.bayern.de/gerichte-und-behoerden/landgericht/muenchen-1/presse/2026/16.php?ref=theseguysknow.io) - [Reuters: German court rules Suno broke copyright rules](https://www.reuters.com/world/german-court-rules-ai-music-firm-suno-broke-copyright-rules-2026-07-31/?ref=theseguysknow.io) - [Suno: The Future of Music](https://suno.com/blog/future-of-music?ref=theseguysknow.io) - [Suno: Warner Music Group partnership](https://suno.com/blog/wmg-partnership?ref=theseguysknow.io) - [Suno: Series D funding announcement](https://suno.com/blog/series-d-announcement?ref=theseguysknow.io)