Sony and UMG sue Suno again, alleging v6 launders earlier copyright infringement

Sony Music and Universal Music Group have filed another copyright complaint against Suno, according to The Verge, this time targeting the AI music company's v6 model. The labels' argument is that v6 infringes even if it was never directly trained on their recordings, because it learned from user outputs of earlier Suno models that were themselves built on unlicensed music pulled from YouTube and other sources.

The laundering theory

The complaint, shared with The Verge, accuses Suno of "model laundering." "Training a 'new' model on the outputs of an infringing model does not eliminate the infringement; it launders it," the labels argue, describing how the value of their expression passes from copied recordings into the older models, from those models into their outputs, and from those outputs into v6. The filing concludes that v6 "is not a fresh start; it is the fruit of the same poisoned tree."

The claim is essentially one of inheritance: infringing data shaped earlier models, those models generated songs for users, and those songs then became training material for the new model. Under that reading, Suno cannot clean up its technology simply by feeding it its own output.

What Suno has said about v6

When v6 launched, Suno's Jack Brody told The Verge that the model was "trained from the ground up, with a new set of data," including "user data," but did not go into detail. Suno later confirmed to Engadget that the dataset includes "creations" from users, and declined to elaborate when asked for comment. Those user creations are exactly what the labels consider tainted, since they were produced by earlier models trained on unlicensed material.

Sony and UMG also stand apart as labels that have not signed a licensing agreement with Suno, which positions them on the litigation path rather than the dealmaking one.

Distillation is also alleged

Sony further alleges that Suno used distillation to build v6, meaning the new model was trained to replicate the results of previous "teacher" models created with infringing data. The complaint contends that "even a model not directly trained on Plaintiffs' recordings is informed by, and benefits from, Suno's retained unauthorized copies." In practical terms, the labels are arguing that nothing short of genuinely starting over would free Suno's models from their origins.

Why it matters

This suit keeps the defining legal fight over AI training data alive and pushes it into new territory. Retraining on synthetic data and distilling knowledge from older teacher models into newer ones are standard techniques across generative AI, not just in music generation. If a court accepts the labels' laundering argument, iterative retraining would offer no escape from earlier infringement, and AI companies would face pressure to prove clean data lineages for every new model generation. If the argument fails, successive model releases could progressively insulate themselves from claims about the original training set.

The case also signals that licensing in AI music is far from settled. While Suno has reportedly pursued agreements with rightsholders, Sony and UMG remain holdouts, betting that litigation can force stricter terms or a definitive ruling on whether training on copyrighted recordings requires permission at all. Whichever way the court leans, the outcome would shape how every company building generative models handles training data, derivatives and the growing pile of AI-generated content now circulating online.