Sony Music and Universal Music Group have sued Suno again. This time they’re going after v6, the AI music startup’s newest model. According to The Verge AI, the labels say v6 still infringes their copyrights even though Suno describes it as a fresh build. Their argument is that v6 learned from the outputs of older Suno models, and those older models were trained on unlicensed music ripped from YouTube and other sources.
The core legal theory is new, and it reaches well beyond music.
🎯 Situation Assessment
The complaint, which The Verge AI has seen, accuses Suno of “model laundering.” The labels spell it out:
“Training a ‘new’ model on the outputs of an infringing model does not eliminate the infringement; it launders it, passing the value of Plaintiffs’ expression from the copied recordings into the tainted models, from those models into their outputs, and from those outputs into v6 … V6 is not a fresh start; it is the fruit of the same poisoned tree.”
Sony and UMG are the notable holdouts here. They never signed a licensing agreement with Suno. Warner Music Group went the other way and settled with Suno last year in exchange for a licensing deal. That leaves two of the three major labels still fighting in court.
📋 What Suno Has Said
Suno’s statements about how v6 was built have been vague so far:
- At launch: Suno’s Jack Brody told The Verge AI that v6 was “trained from the ground up, with a new set of data,” including “user data.” He didn’t go into specifics.
- Follow-up: Spokesperson Rachel Racusen confirmed that the data includes user “creations.” In her words: “v6 was trained on content licensed from our partners, interactions including creations and preference signals from our community, and the accumulated learnings from our team.”
- Still unanswered: Suno hasn’t said whether that training data included audio users uploaded, or outputs generated from uploaded audio.
The third point matters most. If a user uploads a copyrighted track and Suno’s model builds on it, the “user data” label hides where that material really came from.
🧪 The Distillation Charge
Sony goes a step further. It alleges Suno used distillation to build v6.
Distillation is a standard AI technique. You take a large, capable “teacher” model and train a new “student” model to reproduce its outputs. The student ends up absorbing much of what the teacher knows, often at lower cost.
Sony’s claim is that Suno’s teacher models were built on infringing data, so the student inherits the problem. The complaint says “even a model not directly trained on Plaintiffs’ recordings is informed by, and benefits from, Suno’s retained unauthorized copies.”
Put simply, the labels say Suno can’t clean up its models unless it truly starts from scratch.
⚠️ Why This Matters Beyond Music
The labels are testing whether infringement can pass from one model generation to the next. If a court accepts that idea, the consequences go well past Suno:
- Synthetic data gets riskier. Many AI companies train new models on outputs from earlier ones. If the older model’s training data is legally shaky, the newer model could carry the same exposure.
- Distillation comes under legal scrutiny. It’s everywhere in the industry. A ruling that treats a distilled student as tainted by its teacher would make companies audit their whole model lineage.
- “We retrained it” stops working as a defense. Plenty of AI firms have treated a new model version as a clean break. This lawsuit argues it isn’t one.
- Licensing pressure goes up. Warner’s settlement shows one way out. Suits like this one push the remaining holdouts and the AI companies toward the negotiating table.
There’s an interesting twist here. AI labs have argued with each other about distillation for a while, usually over one company copying another’s model outputs. Now rights holders are using the same idea to say that tainted data doesn’t wash out.
🔭 What to Watch
- Suno’s formal response. The Verge AI updated its report with a statement from Suno, and the company’s legal filings will show whether it disputes the distillation claim or the facts about its training data.
- Discovery. If the case continues, Suno may have to explain exactly what went into v6, including whether uploaded audio played a part.
- Settlement signals. Warner settled. Sony and UMG could be using this suit as leverage for a better licensing deal.
- Copycat claims. Expect other rights holders in text, images, and video to borrow the “model laundering” argument if it gets any traction in court.
For AI builders, the practical lesson is clear. Know where your training data came from, including data your own earlier models generated, because you may have to answer for it. The full complaint details and Suno’s statement are in the original report at The Verge AI.