Sony and Universal Music File Second Lawsuit Against Suno
On September 25, 2026, Sony Music Entertainment and Universal Music Group filed a new lawsuit against AI music generator Suno in U.S. federal court. The suit targets Suno’s newly launched V6 model, accusing it of continued copyright infringement. Crucially, this complaint shifts focus: rather than attacking the original training data sources, plaintiffs now target V6’s indirect reliance on outputs from Suno’s prior models.
Key facts at a glance:
- Plaintiffs: Sony Music Entertainment and Universal Music Group (both refused to sign Suno’s licensing agreement)
- Defendant: Suno’s V6 generative model
- Core claim: ‘model laundering’—using outputs from prior (infringing) models to train a successor
- Suno’s response: V6 is ‘trained from the ground up’ using user-submitted and licensed data
The ‘Laundering’ Allegation: Legal Logic Explained
The complaint meticulously outlines what the labels call ‘model laundering’: training a ‘new’ model on outputs generated by previous models that were themselves trained on unlicensed material. As the court document puts it: training a new model on infringing outputs does not eliminate infringement—it ‘launder[s]’ it, passing the value of plaintiffs’ protected expression from the tainted model into the new output.
‘V6 is not a fresh start; it is the fruit of the same poisoned tree.’ This legal metaphor—borrowed from U.S. constitutional law, where evidence derived from illegal searches is inadmissible—captures the labels’ central thesis: proximate causation remains unbroken.
When V6 launched, Suno’s Jack Brody told The Verge it was built on ‘a new set of data’ and ‘trained from the ground up,’ mentioning ‘user data’ but offering no technical details. Spokesperson Rachel Racusen later confirmed V6 incorporates ‘content licensed from our partners, interactions including creations and preference signals from our community, and the accumulated learnings from our team’—without specifying whether uploaded audio or derivative outputs were included.
Distillation and the ‘Teacher Model’ Concern
Sony further alleges Suno employed distillation to train V6: a technique where a smaller ‘student’ model learns to mimic the outputs—or even internal logits—of a larger ‘teacher’ model. If Suno’s teacher models were trained on unlicensed recordings (primarily scraped from YouTube and similar sources), then V6—even if never shown raw copyrighted files—would inherit their learned patterns and biases.
The complaint states bluntly: ‘Even a model not directly trained on Plaintiffs’ recordings is informed by, and benefits from, Suno’s retained unauthorized copies.’ In effect, unless Suno purges all prior data and truly restarts training, the labels argue the legal taint is irremediable.
Industry Licensing Split
| Company | Signed Suno License? | Legal Status |
|---|---|---|
| Sony Music Entertainment | No | Litigation (second round) |
| Universal Music Group | No | Litigation (second round) |
| Other major labels | Yes | Licensing agreement active |
Industry sources note that all major labels except Sony and UMG have reached licensing deals with Suno. Their refusal to negotiate—and decision to litigate—makes them outliers in a rapidly evolving market.
Practical Recommendations
For individual creators using AI music tools for commercial projects: consider switching to Suno’s licensed partner tier—or delay deployment until the court’s ruling on training-data causality becomes clear.
For enterprise applications embedding music generation: demand third-party audit proof that data lineage avoids unlicensed sources, not just contractual indemnity clauses.
Finally: This case forces courts to define how copyright law applies when models learn from other models. How to distinguish between idea and expression, inspiration and copying, transformer training and distillation inheritance—these questions will shape the next decade of AI regulation worldwide.