A new Suno lawsuit filed by Universal Music Group and Sony Music label entities on September 18 asserts copyright claims involving 60,202 sound recordings. The complaint is a set of allegations, not a ruling: its consequential question is whether rebuilding an AI music model with licensed partners can insulate a new system from alleged copying in earlier training generations.

Key takeaways

  • The second federal case identifies 60,202 recordings, far beyond the 560 works in the original 2024 action.
  • The labels plead direct infringement and anti-circumvention claims; Suno had not answered the new complaint when the cited reports published.
  • The dispute now targets model lineage: the labels say v6 inherited value from earlier models, while Suno has described v6 as a clean, licensed rebuild.

Everyone else is reporting the headline count; Lapaas Voice is explaining why the model-lineage argument could matter to every generative-AI company that tries to reset its data strategy.

What the Suno lawsuit actually says

The 45-page complaint was filed in the US District Court for the District of Massachusetts as case 1:26-cv-14275. Twelve Universal and Sony label entities are named as plaintiffs. They allege that Suno copied protected recordings without permission, used them in commercial model training and obtained some audio by circumventing YouTube measures. The filing seeks damages, an injunction and a jury trial.

The number 60,202 comes from the plaintiffs’ list of asserted recordings. It should not be converted into a conclusion that every claim will succeed. The court has not decided infringement, willfulness, fair use, causation or damages, and Suno had not filed its response to this second action when The Next Web, Music Business Worldwide and Bloomberg Law reported the case.

How the second Suno case reaches the courtFour steps distinguish discovery, a denied amendment, a parallel complaint and future merits testing.Discoveryworks identifiedAmendment biddenied without prejudiceParallel case60,202 works assertedMerits stagenot yet decided
The new complaint follows a procedural route; it is not a judgment on liability.

Why the second case exists

The original action dates to June 2024 and covers a much smaller illustrative set. According to the new complaint and independent coverage, the labels later sought to add more than 61,000 works. In August, the judge declined to expand the original case because doing so would disrupt its schedule, while leaving the plaintiffs free to pursue claims in a parallel action.

That procedural history matters. The second filing is not a second liability decision, and the denial of an amendment was not a vindication of either side. It created a separate track for a larger catalogue while preserving the original case’s timetable. Readers should therefore treat the 60,202 count as the scope of pleaded works, not as verified infringement.

The model-lineage argument is the real technology issue

Suno launched v6 on September 9 and said the system was trained from scratch around licensed relationships. The labels’ new complaint argues that a fresh training run does not necessarily break the chain if the new model learned from outputs or preference signals generated by older systems. In their telling, distillation or feedback from earlier models can transfer capabilities that allegedly originated in unlicensed recordings.

That position is contested. Suno has said its licensing partnerships and technical rebuild changed the data foundation of v6. A court would need to examine what data entered each training stage, what outputs or preferences were reused, and whether those steps amount to legally relevant copying. Marketing phrases such as “from scratch” cannot answer those technical and legal questions by themselves.

In plain terms: the Suno lawsuit asks whether an AI company can cleanly reset a model family by changing licensed inputs, or whether alleged value carried through distillation, synthetic data or user-preference signals keeps the old dispute alive. The answer will depend on evidence about the training pipeline, not on the version number.

Scale of works asserted across the two Suno actionsA proportional comparison shows 560 works in the original action and 60,202 in the new complaint.Asserted works, not adjudicated infringementsNew case: 60,202 recordingsOriginal case: 560 worksBar lengths are compressed for readability; the new count is about 108 times larger.
The chart compares pleaded scope and does not imply that liability has been established.

Why licensing deals cut both ways

Universal and Sony point to Suno’s agreements with Warner Music Group, BMG and Believe as evidence that a market exists for licensing recordings or related data. Suno can answer that commercial settlements and forward-looking partnerships do not concede that past training was unlawful. Both propositions can be true: a company may buy certainty for future products while continuing to litigate the legal status of prior conduct.

The contrast with the Suno v6 licensed-model launch is therefore central. That product story described a commercial transition. This court story tests whether the transition changes exposure attached to earlier models. The UMG and ElevenLabs licensing deal shows the parallel market response: rights holders increasingly want negotiated inputs, attribution and payment before a model reaches users.

What the numbers do and do not mean

Reports note that willful infringement can carry statutory damages of up to $150,000 per work, producing a theoretical ceiling above $9 billion if multiplied across 60,202 recordings. That is not a forecast. Courts can award different amounts, reject claims, limit remedies or resolve cases through settlement. The complaint also pleads circumvention remedies, but those too require proof and legal findings.

The useful business signal is not the maximum arithmetic. It is the cost of training-data uncertainty. If model developers cannot produce auditable lineage for copied recordings, synthetic outputs and feedback data, a later licensed model may still inherit litigation risk. Investors and enterprise partners will increasingly ask for dataset records, deletion evidence, model-family boundaries and contractual indemnities.

What happens next

Suno’s answer will establish which factual allegations it admits, denies or says require no response. Discovery could then test the plaintiffs’ fingerprinting work and Suno’s records for training, retention, distillation and preference data. The court may also consider how this parallel case should coordinate with the original action.

Until that happens, the most accurate formulation is narrow: Universal and Sony entities allege infringement involving 60,202 recordings and say v6 remains connected to earlier models; Suno’s contrary account of a licensed rebuild remains unresolved. Any headline that treats the complaint as a verdict gets ahead of the evidence.

Frequently asked questions

What does the new Suno lawsuit allege?

Universal and Sony label entities allege that Suno copied 60,202 sound recordings without permission to train models and also circumvented YouTube access controls. Suno had not answered the new complaint when the cited reports published.

Is the theoretical $9 billion figure an award?

No. It is an arithmetic ceiling derived from the maximum statutory rate cited by reporters, not a damages finding or prediction. Any award would depend on liability, proof and judicial decisions.

Why is Suno v6 part of the dispute?

The labels allege that using outputs or preference data from older models can carry disputed value into v6. Suno has separately said v6 was built from scratch with licensed partners; the court has not resolved that conflict.

Sources

Get the day’s top stories in your inbox

One concise email. No spam, unsubscribe anytime.