What exactly is "model laundering"?
Universal Music Group and Sony Music filed a new copyright complaint against Suno in Boston federal court on September 18, 2026, and it leans on a theory nobody has tested in court before. The labels aren't only repeating their original claim that Suno trained its early models on more than 60,000 unlicensed recordings. They argue Suno's new v6 model, launched September 9 and marketed as built "in partnership" with Warner Music Group, BMG and Believe, still carries the original sin forward, because it was trained in part on the outputs of those earlier, already-infringing Suno models.
The complaint puts it bluntly:
"Training a 'new' model on the outputs of an infringing model does not eliminate the infringement, it launders it."
If that theory holds, licensing a new model with three majors doesn't wash the older model's sins out of its training data, because the new model was never actually separated from what came before it. Statutory damages of up to $150,000 per work could put Suno's exposure above $9 billion if a court agrees.
Why is Canada's rights body suing too, and over which songs?
Universal and Sony aren't alone. Suno is also facing a separate suit from SOCAN, the Canadian performance rights organization representing more than 200,000 songwriters, composers and publishers, filed September 2, 2026. SOCAN's complaint names 150 publicly posted Suno-generated tracks it says reproduce or closely mimic Canadian copyrighted songs, including Avril Lavigne's "Sk8er Boi," Tom Cochrane's "Life Is a Highway," Joni Mitchell's "Both Sides Now" and The Tragically Hip's "Bobcaygeon." SOCAN calls those 150 tracks "the tip of the iceberg" and is seeking statutory damages plus C$10 million in punitive damages.
It's a reminder this fight isn't only a US major-label story. Any rights body with a large enough catalog and the budget to litigate is lining up, and SOCAN naming specific generated tracks that mimic specific songs gives other collecting societies, including the ones that pay out mechanical and performance royalties to house and techno producers, a template for their own suits.
Does the appeals court ruling actually back up the labels?
On September 29, the Third Circuit Court of Appeals handed down the first federal appellate ruling on AI training and fair use, siding with Thomson Reuters against ROSS Intelligence, a now-defunct legal research startup that trained its tool on Westlaw's copyrighted editorial headnotes. The RIAA and the National Music Publishers' Association filed a joint brief backing Thomson Reuters, arguing the ruling "establishes key precedent that training an AI model on copyrighted works to compete with and substitute for those copyrighted works can never be fair use."
But the court's own opinion complicates that reading for Suno. In a footnote, the judges explicitly separated ROSS from generative systems, noting that "unlike the AI models in Bartz and In re: OpenAI, ROSS's AI platform cannot generate original expression." Suno's models generate music, the one capability the court flagged as different. So the ruling is real, it's the first federal appeals court to reject an AI fair-use defense outright, but it also hands Suno's lawyers a specific distinction to argue around rather than a closed door.
What does this mean if you're not signed to a major?
Nobody suing Suno right now represents an independent house or techno label, and that's the uncomfortable part. The model laundering theory, if it survives, would apply just as much to any AI platform that trained once on an unlicensed rip of an independent label's back catalog and then laundered that training through a newer, supposedly clean model. Most bedroom producers and small labels have no idea whether their own catalog, uploaded to Bandcamp, SoundCloud or YouTube over the past decade, has already been scraped into a dataset somewhere, and no realistic way to find out, let alone sue over it.
This case is being fought entirely with major-label money over major-label catalogs. Whatever precedent comes out of it, on model laundering, on fair use, on what counts as a genuinely "clean" retrain, will apply to every artist whose music ends up inside one of these models, whether or not they were ever in the room.



