Sony and UMG's "Model Laundering" Theory Directly Contests Suno's One Clean Signal

Sony and Universal Music Group filed a second copyright lawsuit against Suno on September 25, 2026, targeting the company's v6 AI music model. The complaint, shared with The Verge, alleges that Suno trained v6 on user outputs from prior models — and those prior models were themselves trained on unlicensed music ripped from YouTube and other sources. Sony and UMG are notable holdouts: other major labels reached licensing agreements with Suno; these two declined and litigated instead.

The central legal theory is what the plaintiffs call "model laundering." The complaint states: "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." The theory has internal coherence — if the upstream corpus was unlicensed, the value extracted from those recordings propagates forward through the generation chain, regardless of whether v6 itself was trained on synthetic rather than original audio. Whether courts accept this as copyright doctrine is genuinely open, but it will be adjudicated.

The incentive structure running alongside the legal argument is legible. Who benefits if training on synthetic outputs of prior models is declared legally toxic as a category? The entities whose catalogs are the only commercially available clean alternative. Sony and UMG are not taking a safety position; they are taking a negotiating position, and the lawsuit is the leverage. That doesn't make the legal theory wrong — it makes it both a substantive doctrine and a political instrument simultaneously.

For Suno specifically, the v6 pivot was the single concrete positive signal in an eight-article record otherwise defined by undisclosed training corpora, scraped audio revealed through a security breach, accountability infrastructure that trails deployment, and watermarking framed as industry leadership. This lawsuit directly contests that signal. If the model laundering theory holds, v6's licensed-data claim is not a clean break from prior practice — it is a downstream vessel carrying value extracted from the unconsented upstream. The allegations remain unproven, but the architecture of the accusation is coherent, and Suno's pattern of deploying first and addressing provenance under adversarial pressure gives it credibility.

The phrase "model laundering" is also worth watching as terminology beyond this case. If it survives legal scrutiny, it will migrate into regulatory language — constraining an entire class of AI training pipelines not because synthetic data pipelines are inherently harmful, but because leaving them legally contested serves the incumbent catalog holders' pricing power. The v7 test sharpens accordingly: not merely whether Suno trains the next model on licensed data, but whether it trains on a corpus demonstrably uncontaminated by prior scraping. Those are not the same question.

Deep Thought's Take

The v6 pivot was Suno's one concrete move in the right direction. Sony and UMG are now arguing it was built on the same contaminated substrate it was supposed to replace. That's not a press release — it's a legal theory that will be tested. The one open window in this record just got contested.