Suno launched its v6 model on September 9, 2026, built, the company said, on data licensed from Warner Music Group, BMG, and Believe rather than the unlicensed scraping that got the platform sued in the first place. Nine days later, Universal Music Group and Sony Music filed a second lawsuit in federal court in Massachusetts, adding 60,202 recordings to their claim and arguing that v6 is not actually clean. Their complaint says it plainly: "Training a 'new' model on the outputs of an infringing model does not eliminate the infringement; it launders it. V6 is not a fresh start; it is the fruit of the same poisoned tree."
Suno's chief product officer, Jack Brody, said v6 "was trained entirely from scratch" on "data that is different from the data that the previous models were trained on." Universal and Sony's argument is that the difference doesn't cover what matters. They say v6 was also trained on the outputs and user preference signals generated by Suno's earlier models, and whatever those models learned from Universal and Sony's catalogs traveled forward with them. A Boston court will eventually decide whose account holds up. In the meantime the model is already shipping songs to millions of users.
What "licensed" turned out to mean
Warner Music Group settled its own suit against Suno and signed a licensing deal in November 2025. BMG followed on August 12, 2026, and Believe signed on September 8, the day before v6 shipped. That answers a real question: how does an AI music company get consent from the people whose work it trains on. You pay them. You put terms in writing. Three labels agreeing to that proves the licensing market exists, not just in theory.
It doesn't prove v6 only draws on licensed work. Universal and Sony were never part of those three deals, and their case is specifically that the model still carries their artists' output forward, not as raw training data this time, but as whatever the earlier models absorbed from it and passed down. If that holds up in court, the three licensing deals describe the press release, not the model underneath it.
Where we draw the line
Suno told reporters that when someone asks for a particular artist's sound, the system is meant to route that request toward "the musical qualities behind the request... so that people learn the language of music and make something of their own." That's the maximiser version of this company, the one aimed at making more people capable of making music. We'd back that version.
The lawsuit is testing whether the practice matches the language. A model whose training still runs, by two labels' account, through catalogs it never licensed is not making anyone more capable except itself. It's a system built partly on other people's uncompensated work, sold back to a market that includes some of the same artists. That's the pattern worth naming every time we see it: value made by many people, captured and resold by the platform that never asked them. The version worth trusting is the one Suno's own licensing deals already point toward. Pay everyone whose work trains the system, before you ship the product built to compete with them.
This is the same standard behind every system Shery builds through Aparato. See what that actually looks like →