Introduction

The low-end just got heavier. On Friday, 18 September 2026, Universal Music Group and Sony Music Entertainment filed a second joint lawsuit against Suno in Boston federal court—this time aiming squarely at the platform’s new v6 model suite.

This is not a rehash of the June 2024 fair-use fight over a few hundred works, and it is not the Warner/BMG/Believe partnership beat. The new complaint’s central hook is sharper: labels say v6 is “the fruit of the same poisoned tree”—allegedly built from the outputs, preference data, and distilled “teacher” behavior of earlier models trained on their recordings.

Primary read: Tim Ingham at Music Business Worldwide.

The Poisoned-Tree Claim Against V6

Per MBW’s read of the filing obtained for the report, Universal and Sony accuse Suno of copying 60,202 of their sound recordings without a license to build the models behind its generator.

US copyright law caps willful statutory damages at up to $150,000 per work. At that ceiling, those 60,202 recordings carry a theoretical maximum of just over $9 billion. The labels also seek up to $2,500 per alleged act of circumventing YouTube’s anti-downloading tech—about $150 million if applied once to each of those recordings—plus an injunction against further use of their works and a jury trial.

The timing lands on v6’s launch window. Suno rolled out v6 on 9 September and says it was built in partnership with Warner Music Group, BMG, and Believe. Chief Product Officer Jack Brody told MBW 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 reject that as a clean slate. The complaint alleges v6 was trained using “user interactions” with earlier models—outputs and preference data (Suno generates two tracks per prompt; preference data is which one the user picks)—and via knowledge distillation, where a new “student” model learns to reproduce a predecessor “teacher.” Those teachers, the suit says, include v4, v4.5, v4.5+, v5, v4.5-all, and v5.5, all allegedly trained on the same unlicensed corpus.

The filing’s punchline quote: “Training a ‘new’ model on the outputs of an infringing model does not eliminate the infringement; it launders it… v6 is not a fresh start; it is the fruit of the same poisoned tree.” The complaint also says Suno has never stated it destroyed the recordings it allegedly copied, so v6 “rests on a foundation of unauthorized copies… that Suno continues to hold and exploit.”

Pro Tip

Treat “trained from scratch” marketing and “poisoned tree” litigation as two different mix buses. If you ship with Suno v6 (or any model in a live label fight), keep model version, export date, prompt/project logs, and license/terms screenshots in the session folder. Provenance is the stem that survives a loud courtroom week.

Why a Second Suit—and What It Means for Creators

This filing was triggered by procedure in the first case. UMG and Sony’s original June 2024 suit covers 560 works. They asked to add 61,026 more; on 18 August, Judge F. Dennis Saylor IV refused, saying the extra works would delay a case he wants decided on fair use—and noting the labels could bring them as a separate suit. In discovery on that first case, the labels say they used Audible Magic audio fingerprinting to identify their recordings inside Suno’s training data.

The new complaint also turns Suno’s own deals against its fair-use posture: Warner settlement (November 2025), BMG (12 August), Believe (8 September). Three major rightsholder deals in under a year, the filing argues, show a functioning market for licensing training data—and that Suno itself recognizes permission is required. (Brody’s line that partner revenue share is “not in exchange for training” / “not really about the data” is called “self-serving” in the complaint.)

On market harm, MBW notes the suit cites Deezer’s July 2026 report that more than half of new tracks arriving daily were AI-generated—around 90,000 a day—arguing machine tracks compete for placement, attention, and royalty pools. Separately, the first case still runs: Suno argues training is “quintessential fair use,” admitted in a 1 September filing that it obtained audio from YouTube via YT-DLP, and fact discovery closes 30 September.

Practical moves while Boston sorts the second complaint:

  1. Separate product news from liability theory. Warner/BMG/Believe-backed v6 and this UMG+Sony suit can coexist in the same week—different rightsholders, different theories.
  2. Save the session. Model ID (v6 vs earlier), timestamps, stems, and ToS/license snapshots beat memory when a distributor or counsel asks how a track was made.
  3. Watch disclosure and likeness risk. This suit is about training/distillation of label recordings—not a free pass for uncleared voice clones or “sounds like” marketing.
  4. Don’t over-read the damage ceiling. ~$9B / ~$150M are theoretical statutory maxes reported by MBW, not a verdict.
  5. Follow the primary. Bookmark MBW’s full report and the complaint link published there as the case moves.

Conclusion

UMG and Sony’s second Suno suit is the majors putting a gavel on v6 continuity: not just “did you train on our catalog,” but “does a from-scratch model still carry the juice of earlier alleged infringement through distillation, preferences, and outputs?” Suno says v6 was built from scratch without UMG/Sony data; the labels call that laundering. The theoretical damage ask is enormous; the legal question—poisoned tree vs fresh start—is what Boston will actually weigh.

For generative producers, keep the pocket clean: know which model cut your track, keep receipts, and treat label partnership headlines and infringement headlines as separate faders on the same board. Heavy impact still cuts through. Contested provenance is what gets soloed in court.

—Anabel ♡