The $9 Billion Question: Universal and Sony Sue Suno Again, Calling v6 'the Fruit of the Same Poisoned Tree'

On September 9, Suno shipped what looked like the end of the AI music wars: a v6 model family trained in partnership with Warner Music Group, BMG and Believe, with revenue sharing live from launch day. Nine days later, the other two majors replied the way labels historically reply — with a 45-page complaint.

Universal Music Group and Sony Music Entertainment filed a new lawsuit against Suno on Friday, September 18, in the U.S. District Court for the District of Massachusetts. It is their second case against the AI music company, and it takes direct aim at the very model Suno built to escape the first one. The complaint accuses Suno of copying 60,202 of the labels’ sound recordings without a license and using them — directly and indirectly — to build the models that power its service today. At the statutory ceiling of $150,000 per willfully infringed work, that count carries a theoretical maximum of just over $9 billion.

The poisoned tree

The heart of the new suit is a single, elegant legal theory: that v6 is not a clean restart but a laundering operation. Suno’s chief product officer Jack Brody said at launch that v6 “was trained entirely from scratch, from the ground up” on data that “doesn’t include data from Universal or Sony.” The labels do not dispute that the original recordings went into a different generation of models. They argue it doesn’t matter.

“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,” the complaint states. “V6 is not a fresh start; it is the fruit of the same poisoned tree.”

Two technical mechanisms get named. The first is Suno’s own admission that v6 was trained on “user interactions” with previous models. Every time Suno generates two tracks per prompt and records which one the user prefers, that preference signal is supervision data — and, the labels argue, every such signal has so far been “a judgment about audio outputs generated using Suno’s unlicensed models,” synthetic audio that embodies expressive features derived from the plaintiffs’ recordings.

The second is knowledge distillation, described in the filing as a process “whereby a new ‘student’ model is trained to reproduce the learned behavior of a predecessor ‘teacher’ model.” Here the teachers are Suno’s prior models — v4, v4.5, v4.5+, v5, v4.5-all and v5.5, all named in the complaint — and the capabilities they transmit, “how to compose, arrange, and render convincing imitations of human-created music across genres and styles,” were allegedly derived from the unlicensed corpus. In substance, the filing argues, “v6’s training corpus embodies the accumulated learnings of models built on Plaintiffs’ works.”

The complaint also notes that Suno has never said it destroyed the recordings it copied. “V6 thus rests on a foundation of unauthorized copies of Plaintiffs’ works that Suno continues to hold and exploit,” it reads.

Turning Suno’s deals against it

The sharpest rhetorical move in the filing is that it uses Suno’s licensing momentum as evidence against Suno. Warner settled with Suno in November 2025, BMG signed on August 12, and Believe announced its partnership on September 8. “Three agreements with three major rightsholders in less than a year confirm that a functioning market exists for licensing sound recordings to train generative AI models,” the complaint argues — undercutting the fair-use defense’s market-harm analysis, since Suno’s own conduct demonstrates that permission was obtainable.

The labels quote Brody describing the revenue share as “not in exchange for training” and the deals as “not really about the data” — a characterization the filing calls “self-serving.”

On market harm, the complaint reaches for a number that has unsettled the industry all year: Deezer reported in July 2026 that more than half of the new tracks arriving on its platform each day are AI-generated, roughly 90,000 daily. “Every machine-generated track delivered to a streaming service competes with the Copyrighted Recordings for placement, for discovery, for finite listener attention, and for a share of the fixed royalty pools from which genuine artists and rightsholders are paid,” the filing states. “Dilution at this scale and velocity is not just a future harm; it is a present, measured, and accelerating one.”

Even Suno CEO Mikey Shulman’s old remark that “it’s not really enjoyable to make music now” makes an appearance — cited as proof that “the time, practice, and skill Mr. Shulman derides are exactly what copyright law exists to encourage and reward.”

Why a second lawsuit

The timing was not purely tactical. In the first UMG/Sony case, filed in June 2024 alongside Warner, the labels recently asked to add 61,026 recordings to the 560 works already in suit. Judge F. Dennis Saylor IV refused on August 18, ruling that the expanded case would delay a ruling he wants decided on the fair-use question, while noting the labels were free to bring the additional works in a separate action. This new complaint is that action. The labels say they pinpointed their works inside Suno’s training data during discovery, using audio-fingerprinting company Audible Magic — and they emphasize that the 60,202 count “remain[s] only a small portion” of the total.

A separate circumvention claim adds another layer: the labels seek up to $2,500 per act of defeating YouTube’s anti-downloading technology, which applied across the same recordings would add roughly $150 million. Suno admitted in a September 1 filing that it obtained audio from YouTube using YT-DLP. The labels also ask for an order stopping any further use of their recordings, and they’ve demanded a jury trial. Fact discovery in the first case closes September 30.

Suno’s answer

A Suno spokesperson called the claims “fundamentally flawed on both the facts and the law,” and framed the platform’s purpose as helping people “learn the language of music and make something of their own.” The statement points to v6’s launch partnerships: “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.”

That sentence is, in miniature, the entire dispute. Everything Suno lists is either licensed, user-generated, or learned — and the labels’ entire theory is that the last two categories are downstream of the unlicensed first act. Suno has also said the older, unlicensed models would no longer be supported, which the complaint dismisses as beside the point while the derived models remain in production.

The bigger fight

Whichever way Judge Saylor rules on fair use in case one, this second suit ensures the “tainted lineage” question survives it. The theory at stake — that a model trained on the outputs of an infringing model inherits the infringement — has no settled precedent, and it generalizes far past music. Every AI lab that has retrained a cleaner model on data generated by an earlier, legally contested one is watching this docket. If the labels’ laundering theory wins, “we deleted the old checkpoints” stops being a defense; if it loses, model succession becomes a reliable statutory washing machine.

For now, the industry’s split reality continues: Warner, BMG and Believe collect royalties from Suno v6 while Universal and Sony pursue it toward a potential nine-figure — or nine-billion-dollar — judgment. The AI music market has both a licensed lane and a war zone, and this week they are running through the same company.