The Spotify/UMG AI deal is a blueprint publishing could have written. Instead, it will try hard to not even read it.

Back in February, Spotify co-CEO Gustav Söderström told analysts that the platform’s AI remix technology was already built and ready to deploy, but that “the absence of a rights framework” was holding things up.

Publishing looked the other way. Not even HarperCollins, by far the most engaged with AI opportunity, dared go that far!

Söderström was saying something publishing has not yet found the vocabulary to say out loud, even if it has privately reached the same conclusion. The obstacle was never technical. It was contractual. And the only people who could remove it were the rights-holders.

Three months later, Spotify and Universal Music Group have announced what both companies called a landmark licensing deal. The music industry’s stock market rewarded Spotify with a 16% share price jump.

Premium subscribers will soon be able to create AI-generated covers and remixes of songs by participating artists. The feature will launch as a paid add-on. Revenue will flow back to the artists and songwriters who opt in. The framework rests on three pillars: consent, credit, and compensation.

The Walled Garden is the Point, not the Flaw

The deal is not simply a new feature. It is an architectural decision.

In late 2025, UMG settled its copyright infringement lawsuit against AI music platform Udio and agreed to collaborate on a licensed AI music creation service – one built around a “walled garden” model in which AI-generated content cannot be downloaded or distributed outside the platform on which it was created.

The Spotify deal extends that logic. The AI-generated remixes live inside Spotify. They play inside Spotify. The value they generate is captured inside Spotify, and a portion of it flows back to rights-holders.

This is not accidental. MBW (Music Business Worldwide) has reported on a patent portfolio linked to UMG targeting AI music derivative infrastructure, including systems built around artist approval, automated revenue distribution, and digital watermarking. The architecture was being designed before the announcement. The rights framework, once secured, slotted into the infrastructure already waiting for it.

Publishing should read that sequence carefully. The music industry did not wait for consensus. It built the system, then signed the deal that activated it.

Publishing has been here before, and walked away

In 2013, Amazon launched Kindle Worlds- a licensed platform allowing fan fiction writers to publish derivative works set in established fictional universes, with revenue shared between Amazon, the original rights-holders, and the fan authors. It ran for five years and was quietly shut down in 2018.

Kindle Worlds was not killed by lack of reader appetite. Fan fiction has a vast, demonstrably engaged audience; Archive of Our Own alone now hosts over twelve million works. What killed Kindle Worlds was the refusal of major publishers to participate. Without the Big Five catalogues, the platform’s pool of available IP was too shallow to sustain the ecosystem. Amazon had the platform, the infrastructure, and the audience. It did not have the rights.

A decade later, publishing faces the same structural question under higher stakes and with less time to deliberate. The difference is that the music industry has now demonstrated not just that the model works but precisely how to build it. The Spotify/UMG deal is not a thought experiment. It is a working blueprint.

What the blueprint actually shows

The mechanics matter, so pay attention before the Luddite Fringe start their disinformation campaign.

Participation is opt-in – artists and songwriters choose whether to make their catalogues available. All Spotify users can listen to the created tracks, but creating them is a paid premium feature. The revenue model creates additional income on top of existing royalties rather than cannibalising them. The consent, credit, and compensation formula is not window dressing; it is the structural condition that makes the whole arrangement legitimate rather than extractive.

For publishing, the translation is straightforward in principle if not in practice. An author whose backlist is licensed to a major platform could, under an equivalent framework, allow readers to generate licensed continuations, alternative endings, or genre-transpositions of their work – with each interaction generating a micropayment, an attribution record, and a revenue share.

The technology to do this exists.

The demand to do this demonstrably exists, as two decades of fan fiction culture confirms.

What does not yet exist is the rights framework.

And unlike in music, where three major labels plus the independent sector’s Merlin aggregator constitute a manageable number of negotiating parties, publishing’s rights landscape is vastly more fragmented – and critically, far more author-retentive, with many writers holding digital and derivative rights never formally assigned to their publishers.

That fragmentation is a real obstacle. It is not, however, the primary one.

The primary obstacle is political, not structural

The primary obstacle is the anti-AI faction that has made it effectively impossible for major publishers to explore licensed AI derivative frameworks without being accused of betraying their authors.

The position is held with genuine conviction by many, disingenuously by many more, but whether heartfelt or just chasing soundbites, its practical effect has been to ensure that the industry’s default posture is defensive, reactive, and framed around resistance rather than architecture.

Publishers Weekly’s front page this week – published alongside the Spotify/UMG announcement – led with “AAP Partners with Vermillio to Combat AI-generated Audiobook Piracy.“

That is publishing’s characteristic mode: a consortium assembled to fight something that a different consortium, in a parallel industry, just turned into a revenue stream and a 16% stock price gain.

The fence these publishers have impaled themselves on is not neutral ground. Every month spent so-impaled is a month during which Amazon, Apple, and Spotify are quietly building derivative content infrastructure.

When that infrastructure is complete, publishers will be invited to participate on terms they did not negotiate – or were not invited at all, because the platforms will simply approach authors directly. The anti-AI stance, held long enough, does not protect authors from AI derivatives. It guarantees that authors will face those derivatives without publishers in the room.

The exclusion that nobody is discussing

The Spotify/UMG deal has attracted a handful (one finger’s worth) of positive commentary asking whether it offers a blueprint for the creative industries. Camilla Silfvenius, take a bow. That, as Dr. Alfred Lanning’s hologram tells Detective Spooner in I, Robot, is the right question,

The more uncomfortable question is: a blueprint for whom?

The consent, credit, and compensation framework functions only within the formal rights infrastructure – the system of registered copyrights, contracted royalty flows, and institutional rights-management relationships that constitutes the global music and publishing economy.

But here’s what matters beyond the western publishing bubble: That infrastructure covers, at best, a fraction of the world’s creators. Writers and musicians across much of South and Southeast Asia, Oceania, Latin America, and sub-Saharan Africa exist largely outside it – not by choice, but because the infrastructure was built for and by the Anglo-American publishing and music establishment, and was never seriously extended elsewhere.

This is not a peripheral concern. It is a structural fact about who gets to participate in the value the AI derivatives market will generate. When Söderström says “it hasn’t been possible for existing creators to participate because there was no legal licensing framework,” the unspoken qualifier is: existing creators within the formal western system. For the majority of the world’s creative output, there is no legal licensing framework to build on – and no consortium with the institutional weight to negotiate one.

The Spotify/UMG deal is a significant step. But its geography is the same as every previous significant step in the digital content economy: it begins and ends in the markets where the rights infrastructure already exists. (Although Spotify is uniquely positioned to extend that reach.)

What Publishing Needs To Do (But Won’t)

Music did not arrive at the Spotify/UMG deal through consensus or goodwill. It arrived there because the labels had already lived through Napster, through the early streaming wars, through the litigation against Suno and Udio – and had learned, painfully, that refusing the platform deal does not make the platform disappear. It makes the platform more powerful, and the rights-holder less relevant.

Publishing has not yet had that forcing moment. But the logic of the situation is identical. AI-generated derivative content based on existing books is already being created, already circulating, already building audience. The question is not whether it will exist. The question is whether publishers and authors will be inside the framework that monetises it, benefitting as common sense suggests they should, or outside it watching the value flow elsewhere as the Luddite Fringe would prefer.

The Spotify/UMG deal shows that inside is possible. It requires platform partners willing to build the architecture, rights-holders willing to engage rather than litigate (you can feel the Luddite Fringe quaking at the very notion), and the institutional courage to opt in before the market makes the decision for us.

Publishing has the first condition. Several platforms, Spotify itself included given its expanding audiobooks presence, have strong incentives to develop the equivalent framework for books. What remains missing is the second and third – and with every month the Luddite Fringe anti-AI chorus holds the industry on the fence, the window for shaping the terms narrows.

Publishing needs to ease itself off the fence, even at the cost of a few painful splinters, and a) start engaging with its creatives about what AI brings to the creative table, and b) call out the self-serving soundbites and short-term interests of the so-called industry advocacy groups like the UK’s Publishers Association and the Society of Authors for what they are.

This post first appeared in the TNPS LinkedIn Analysis newsletter.