Suno has launched v6, a new generation of AI music models developed with Warner Music Group, BMG and Believe. The company says its previous models will be retired as it moves entirely onto the new generation, while participating rightsholders are beginning to supply licensed music for the new system. :contentReference[oaicite:0]{index=0}
That makes v6 more interesting than an ordinary model upgrade.
For much of the generative-AI fight, permission has been something argued about after the technology was already built. Models were trained, products were launched, lawsuits followed, and licensing negotiations came later. Suno's new direction begins to reverse that sequence.
The important change isn't simply that AI music is becoming licensed. It's that permission is starting to become part of how the product is designed.
From licensing the catalog to designing the relationship
Suno's relationship with the music business has changed dramatically. Warner Music Group settled its copyright lawsuit against the company last year and entered a partnership with Suno. BMG followed with an agreement covering recorded music and publishing repertoire, and Believe announced its own partnership this week covering participating Believe and TuneCore artists and labels.
Those agreements are now showing up in the technology.
Suno Chief Product Officer Jack Brody told Music Business Worldwide that v6 was trained from scratch, with licensed training repertoire managed by participating rightsholders. Suno also says the new models were developed with its industry partners rather than simply being followed by licensing agreements once development was complete.
That distinction matters.
A license can give a company permission to use something. But an AI system capable of generating new media eventually needs to understand more than whether a pile of training data was legally acquired. It needs relationships around participation, authority, permitted uses, resulting works and the people connected to them.
Suno's next phase makes that even clearer.
Opt-in changes the architecture
Suno says it is developing artist-specific experiences in which artists can choose to participate and be paid when they do. Believe's agreement similarly says participating artists and labels can opt in to new licensed products the companies develop together.
That moves the problem somewhere much more interesting.
Once an artist can opt into a particular AI experience, the system needs to know more than whether Suno has a broad relationship with a label or distributor. Someone has to establish which artist is participating, which repertoire or rights are involved, what has been authorized, what the resulting experience is allowed to do and who has authority to make those decisions.
Those aren't merely licensing questions anymore. They are infrastructure questions.
And they become even more complicated when generated media moves beyond completely original output.
Brody told MBW that Suno currently considers its generated songs original creations, but said some ownership rights and other details will differ as the company introduces derivative experiences involving artists who opt in. :contentReference[oaicite:4]{index=4}
That's a significant boundary.
The moment an AI product moves from "make me a new song" toward "let me create something with this artist," the relationship between identity, permission, source material and the resulting work becomes part of the product.
Permission needs to become machine-readable
The music business already knows how to grant permission. It has spent decades doing it through contracts, licenses, rights databases, collection systems and commercial agreements.
Generative AI changes the speed at which those permissions may need to be understood.
A system can generate a piece of media in seconds. If artist participation is going to be meaningful at that speed, authorization cannot exist only as language buried inside agreements between companies. Eventually, the product itself needs a reliable way to understand who is participating, what they control, what they have permitted and what relationships should travel into the resulting work.
That is where this becomes relevant to Certifyd.
Certifyd Core approaches identity, works, rights relationships and derivatives as things that can be expressed and verified by infrastructure rather than recreated independently inside every application. The point isn't to replace licensing agreements. It's to give software a way to understand the relationships those agreements depend on.
An artist could establish identity. A work could have an authoritative relationship to that identity. Participants and permissions could be attached to the work. A derivative could retain a relationship to its source. Different applications could then interact with those records without requiring the creator's identity and authority to begin again inside every company database.
Suno does not need to become Certifyd for this direction to matter. In fact, the more interesting possibility is the opposite: AI companies, distributors, labels, creator tools and other services should be able to interact with creator-controlled records rather than each building another closed version of the creator.
Licensed AI is only the beginning
Suno v6 is important because it suggests the music industry's AI debate may be moving into a different phase.
The first argument was largely about whether AI companies had permission to train their systems at all. That argument is far from over: Universal Music Group and Sony Music Entertainment remain in litigation with Suno over its earlier practices.
But alongside those fights, another model is emerging.
Warner, BMG and Believe are working with Suno. Participating repertoire is entering licensed models. Revenue is being shared with rightsholders. Artist-specific opt-in experiences are being developed. And Suno is already acknowledging that derivative experiences will introduce different rights and ownership considerations.
The question is therefore getting bigger than whether AI companies can obtain licenses.
The next question is whether permission itself can become part of the architecture.
Because once machines can create media instantly, the infrastructure that tells them who authorized what needs to move just as fast.