Sony and Universal Sue Suno Again: The AI Music Copyright War Escalates

The Major Labels Strike Again

Sony Music and Universal Music Group (UMG) have filed a new round of lawsuits against AI music generation company Suno. This isn't the first time the two record label giants have taken legal action against Suno — and this time, the focus is squarely on Suno's latest v6 model.

According to the original report, both labels allege that Suno's v6 model continues to infringe on their copyrights. The reasoning is straightforward: v6 was trained on user-generated outputs from earlier models, and those earlier models were themselves trained on music scraped without authorization from platforms like YouTube. In other words, even though Suno has updated its model version, the copyright contamination has been carried forward through the training pipeline.

The "Transmission" Logic of Copyright Contamination

The most notable technical argument in this lawsuit is what might be called the "data transmission" problem. Suno may have assumed that retraining on a new-generation model or user-generated content could somehow "launder" the legal risks associated with the unauthorized music used in earlier training. But Sony and UMG argue precisely the opposite — if a model's training material comes from the outputs of a previous model, and that previous model learned from stolen music, then the "DNA" of infringement doesn't disappear just because the version number changed.

If this argument holds up in court, it will have far-reaching implications for the entire generative AI industry. It would mean companies cannot simply break their legal connection to original infringing data by "training a new model on the outputs of their old one." For AI companies that got their start using web-scraped data, this sets a particularly dangerous precedent.

This argument touches on the legal concept of "derivative works" under U.S. copyright law. Copyright law stipulates that derivative works based on copyrighted material constitute infringement if created without authorization. The labels' claim is that the musical styles, melodic patterns, and harmonic structures that AI models "learn" through training are essentially copies and distillations of original copyrighted works — meaning the model weights themselves carry an infringing nature. When those weights are used to generate new outputs, and those outputs are then used to train the next generation of models, the infringing content persists in a form that's difficult to detect. This mirrors the "money laundering" analogy in traditional copyright law: mixing illegal proceeds into legitimate assets doesn't make them legal. Academic debate over whether AI model training constitutes "fair use" remains heated, but if the "transmission" theory receives judicial validation, it will significantly raise the compliance bar for AI companies.

Why Sony and UMG Are the Holdouts

The report specifically notes that Sony and UMG are notable "holdouts" in the industry — they have not signed licensing agreements with Suno. This detail reveals a divide within the music industry over how to approach AI training licenses.

Some rights holders may have already chosen to negotiate with AI companies and reach paid licensing deals, treating AI training as a new revenue stream. Sony and UMG, however, have taken a hardline confrontational approach, using litigation to establish stronger bargaining leverage — or even to fundamentally define the legal boundaries of using copyrighted music for AI training.

This "litigation as negotiation" strategy is not uncommon in the content industry. By applying sustained pressure, the labels hope to gain stronger influence over licensing prices, data usage scope, and revenue sharing from AI-generated music.

What This Lawsuit Means for the Industry

The clash between generative AI and copyright has spread from images and text into the music domain, and Suno finds itself at the center of the storm. This new lawsuit once again highlights a core unresolved question: does training an AI model on unauthorized data constitute "fair use," or is it clear copyright infringement?

U.S. courts have yet to deliver a unified answer, and rulings in different cases could shape entirely different industry landscapes. If courts side with the labels, AI music companies will face substantial licensing costs and may need to fundamentally reconsider their training data sources. Conversely, if Suno is able to hold its ground, it could provide a degree of legal cover for how the broader industry uses data.

Regardless of the outcome, this protracted battle between Sony, UMG, and Suno will serve as a crucial bellwether for the evolving relationship between AI and the creative industries. For practitioners focused on AI compliance and copyright issues, the case is one to watch closely.

The "fair use" doctrine under U.S. copyright law is typically evaluated along four factors: the purpose and character of the use (whether it is transformative), the nature of the original work, the amount and substantiality of the portion used, and the effect on the potential market for the original work. AI companies typically argue that training constitutes "transformative use" — extracting statistical patterns from large datasets rather than directly copying content. The labels counter that AI music generation tools compete directly with human musicians in the marketplace, and that users can use them to bypass purchasing legitimate licenses, making the market harm self-evident. Several similar cases involving image-generation AI tools (such as Stability AI and Midjourney) are currently being heard in U.S. federal courts, and their rulings will interact with music copyright cases to collectively define the legal boundaries of training data use in generative AI.

Takeaway

At its core, this lawsuit isn't simply about "copying" — it revolves around the more complex legal question of training data legitimacy and "copyright transmission." It serves as a reminder to all AI companies: model versions can be updated, but the original sins baked into training data may not be so easily erased. Until licensing rules are clarified, the compliance of data sources will remain a Sword of Damocles hanging over every AI company in the business.