From Defendant to Revenue Share: Suno's v6 Was Retrained on Licensed Warner, BMG and Believe Catalogs, and Started Paying Rightsholders on Day One

Suno released its v6 family (v6, v6-Wild, v6-mini) on September 9 and retired all previous models the same day. Chief Product Officer Jack Brody said the training set incorporates verified catalogs from Warner Music Group, BMG and Believe Music under a revenue-sharing agreement, and that from launch day "a portion of our revenue will be shared with our partners, and those rightsholders are responsible for figuring out how to distribute that capital." Per the company, v6 was trained from scratch on licensed Warner and BMG data plus user-generated content, and did not use the dataset behind its older model generations. WMG CEO Robert Kyncl said artists and songwriters will share in revenue generated by Suno's more than 2 million paying subscribers. v6 and v6-Wild go to Pro and Premier subscribers ($8 and $24 a month in the US), while v6-mini stays free for everyone. Suno also committed to audio watermarking and fingerprinting and capped how many songs subscribers can download monthly. Undisclosed: the dataset's size, which artists it covers, and the exact distribution terms.

Both Tracks Are Running at Once

The music industry's posture toward generative AI has split in two this year. This site covered Sony Music Publishing and Warner Chappell suing Anthropic on September 1 — that is the sue-if-you-won't-negotiate track. Suno is the other one: Warner began working with it last November, after Universal, Sony and Warner jointly filed copyright suits; BMG reached its agreement in August; Believe joined on Tuesday. Going from defendant to partner in under a year is a bigger story than a model update. It says the majors have done the arithmetic: litigation produces damages, but not a recurring revenue stream, while turning a catalog into licensed training data produces the latter.

"Retrained From Scratch" Is the Technical Precondition

The line easiest to skip past, and the most important for judging the legal position, is that v6 was trained from scratch on licensed data plus user-generated content, and **did not use** the dataset behind previous model generations. That is also why the older models were retired the same day — two model families with different data provenance would otherwise be running in production under two different legal statuses. It also draws a boundary in the other direction: v6's licensing is not retroactive. Work generated with the earlier models is in exactly the copyright position it was in before this launch. One more qualification deserves a careful read: not every song controlled by Warner, BMG or Believe automatically enters the training pool — participation depends on the relevant agreements, rights and the artist's own choice. Suno also says opt-in experiences are coming, where individual artists can choose to take part in AI-generated projects and be compensated.

The Revenue Share Is Real; the Percentage Isn't Stated

"Sharing starts on launch day" is a checkable commitment, but three blanks follow it: how large the dataset is, which artists it covers, and on what terms the money is split — Brody disclosed none of the three. Kyncl's framing is that artists and songwriters will share in revenue from more than 2 million paying subscribers, without saying which portion of revenue or at what rate. For an artist, "you will be paid" and "you will be paid this much" are two different questions, and the answer to the second sits with the rightsholders — Brody said explicitly that distributing the money is their job. The split between labels and artists has never been automatically fair, and adding a layer to the chain is a reason to press harder on transparency, not less. For creators actually using Suno, two things change concretely: v6 and v6-Wild require a Pro or Premier subscription ($8 and $24 a month in the US) while v6-mini stays free, and monthly downloads are now capped, alongside audio watermarking and fingerprinting. Together that tightens the generate-and-walk-away pattern and makes output more traceable.

via: Music Business Worldwide, Variety, Digital Music News