UMG and Sony Music File $9 Billion Second Lawsuit Against Suno Over 'Model Laundering' in AI Music v6

By Hector Herrera | September 28, 2026 | Creative

Universal Music Group and Sony Music filed a second copyright infringement lawsuit against AI music generator Suno in Boston federal court on September 18, alleging that Suno's v6 model was trained on 60,202 copyrighted recordings without authorization and introducing a legal theory the complaint calls "model laundering" — using outputs from earlier infringing models as training data for a new model release. The theoretical statutory damages exposure exceeds $9 billion.

The First Lawsuit and What Changed

The labels filed their first lawsuit against Suno in June 2024, alleging infringement in Suno's earlier models. That case is still pending. The September 2026 second complaint is not an amendment — it is a separate action targeting Suno's v6 model specifically and advancing a novel infringement theory that did not appear in the original filing.

According to The Hollywood Reporter, the v6 complaint alleges that:

• Suno's v6 model was trained on 60,202 copyrighted sound recordings owned by UMG, Sony, and their affiliated labels
• The recordings were used without authorization or compensation
• Suno did not simply repeat the original infringement — it trained v6 partly on outputs generated by earlier infringing models, creating what the complaint calls "model laundering"
• At $150,000 per infringed recording (the maximum statutory damages under the Copyright Act), the theoretical exposure exceeds $9 billion

Suno called the claims "fundamentally flawed," asserting that v6 was trained on licensed content and user-generated interactions. The company did not address the model laundering theory specifically in its initial public response.

What "Model Laundering" Actually Means

Model laundering is not yet a recognized legal term — the labels are asking the court to accept it as a valid framework for copyright liability. The theory works like this:

1. Suno builds Model v1 using unlicensed copyrighted recordings (the alleged original infringement)
2. v1 generates outputs — music that may not directly reproduce the training recordings but is statistically shaped by them
3. Suno uses those v1 outputs, plus potentially other data, to train Model v6
4. Under the labels' theory, the infringement embedded in v1 was "laundered" into v6 through the intermediate training step — meaning v6 inherits the copyright taint even if its direct training data appears clean

This theory, if accepted by a court, would have profound implications for the AI industry well beyond music. Every AI company that has iterated through model versions using outputs from prior models — which is standard practice in machine learning — could face downstream liability for infringement that occurred in earlier versions, even after attempting remediation.

The legal question is whether courts will treat model outputs as "tainted" by the copyright status of training data, and whether that taint can propagate through successive training runs. There is no controlling precedent on this specific question.

The Statutory Damages Math

The $9 billion figure comes from the Copyright Act's provision for statutory damages — a per-work penalty that does not require proof of actual economic harm. At the maximum rate of $150,000 per willful infringement, 60,202 recordings produces potential exposure of $9.03 billion.

Courts rarely award maximum statutory damages across large catalogs. In practice, judges exercise discretion to set per-work damages at levels they consider proportionate. But statutory damages exist precisely to make copyright infringement economically irrational for defendants who would otherwise calculate that expected damages are lower than licensing costs. The labels are not necessarily expecting a $9 billion judgment — they are creating a negotiating posture and litigation risk that makes settlement attractive for Suno.

Why Suno, and Why Again

Suno is the largest commercially available AI music generation platform by user base. Its business model — generating original-sounding music on demand from text prompts — is in direct competition with the labels' own streaming revenue and potentially with the entire market for licensed production music, stock music, and commercial sync licensing.

The first lawsuit was filed alongside an identical action against Udio, another AI music generator. The music industry's strategy appears to be using Suno as a test case to establish legal precedent before AI music generation scales to a point where the market damage is harder to reverse.

The second lawsuit's timing — filed after Suno released v6 as a materially improved product — signals that the labels are not waiting for the first case to resolve before challenging each new model version. If that strategy succeeds, AI music companies will face serial litigation with each model release, creating a legal overhead that functions as a licensing negotiation lever even without a final judgment.

What the Outcome Could Mean

If the model laundering theory is accepted:

• AI companies will face liability for iterative training — not just for initial training data decisions, but for any subsequent model that was shaped by outputs from an infringing predecessor
• Remediation becomes legally risky: a company that attempts to fix copyright problems by retraining on cleaner data could inadvertently "launder" the original infringement into the new model

If the theory is rejected:

• AI companies retain the ability to iterate models without inheriting infringement liability across versions
• The labels will need to challenge each model release as a fresh infringement rather than tracing liability through model lineages

The case will not resolve quickly. Federal copyright litigation typically takes two to four years to reach judgment. In the interim, the music industry and AI music companies are in a negotiating environment shaped by the outcome of these cases — which is exactly where the labels want to be.

What to Watch

The court's treatment of the model laundering theory in early procedural rulings — particularly whether it survives a motion to dismiss — will be the first legal signal of whether the theory has viability. Separately, watch whether Suno moves toward licensing agreements with major labels before the litigation reaches discovery. Several other AI music companies have already signed blanket licensing deals with labels; Suno has so far declined that path.