The most important AI music platform of the next decade may not be the one that makes the best song.
It may be the one that knows who is allowed to make the song in the first place.
I. The wrong question
The public conversation about AI music has been almost entirely about output.
Is it good? Is it soulless? Is it cheating? Can a machine write a melody that moves you? Can it produce a track indistinguishable from a session in a $5,000-a-day studio? Can it clone a voice so precisely that even the person whose voice it is can't tell?
These are interesting questions. They are not the important ones.
The important question is more administrative, more procedural, and considerably less sexy: Who approved this? Who owns the input? Who gets credited? Who gets paid?
That is the quiet crisis living inside the AI music boom — the gap between what the technology can do and what the infrastructure currently knows how to handle. The generators are miles ahead of the rights systems. The creation layer has been rebuilt from scratch. The accountability layer has not.
And the people building in that gap — the founders, the operators, the IP attorneys, the platform architects who understand that the next version of the music business will be defined by whoever controls the metadata, the rights rails, and the consent infrastructure — they are building something that most of the conversation is not paying attention to yet.
I have been building there. And I want to explain what it looks like from the inside.
II. What the fix has to do
The system the AI music industry actually needs is not a streaming service. It is not a music generator. It is not a search tool or a sync marketplace or an AI model trained on a catalog of samples.
It is rights infrastructure for the AI music era — the kind of system that answers, at the moment of creation and before the moment of release, the question that the entire industry is currently answering after the fact, if it answers it at all: Is this allowed to exist?
I've been calling this category of system pre-flight rights infrastructure. The argument has three layers.
A governed creation workspace. On the surface, it would look like any other AI music tool — intuitive, fast, designed to lower the friction between idea and output. The critical difference is the constraint underneath: every source asset a user works with must be owned by them, licensed, opted in, or cleared. The rule is almost constitutional: ownable or not allowed. There is no workaround. The gate is the product.
A release-readiness layer. Take finished works and prepare them for release. Check AI disclosure status. Read and validate metadata. Confirm ownership, splits, and cover licensing. Flag sync restrictions and brand permissions. Surface evidence of human authorship. Create the audit record that the downstream ecosystem — distributors, streaming platforms, sync libraries, labels, PROs — will increasingly require before they touch a piece of music.
An enterprise gate. This is the infrastructure play. A system that distributors, streaming services, sync platforms, and rights organizations can integrate to score, hold, quarantine, reject, or escalate AI-adjacent uploads before they enter the catalog. It is not a filter for quality. It is a filter for legitimacy.
That last distinction matters enormously. Most digital systems are built around post-hoc enforcement. A track goes live, a content match triggers, a rights dispute follows, a revenue hold lands, a takedown occurs. The cycle is expensive for everyone, and the artists and small operators who can least afford to fight it absorb the most damage.
The argument is structural: AI-era music needs a pre-flight system. Not a vibe check. A gate. Not a correction after the fact, but a verification before the fact.
III. Why the timing is not accidental
The scale of the problem became undeniable in 2024.
The RIAA filed high-profile lawsuits against Suno and Udio, two of the most visible AI music companies, alleging unauthorized use of copyrighted recordings to train their generative models. Those suits did not resolve the legal questions around AI music. They clarified the battlefield. They established that the industry is paying attention, that litigation is a live tool, and that the question of what goes into an AI model is going to be litigated — publicly, expensively, and for years — if the infrastructure around consent and provenance is not built now.
At the same time, Spotify executives have publicly referenced the scale of daily uploads to the platform — figures in the range of tens of thousands of tracks per day, with some estimates significantly higher. Even setting aside the precise number, the directional truth is obvious: music supply has already exceeded the capacity for human-scale review at every distribution point in the system. The review problem is not going to be solved by hiring more people. It is going to be solved by building better systems.
Meanwhile, the Recording Academy, the DDEX standards body, the major publishers, and a growing number of independent rights organizations are actively building frameworks for AI disclosure, metadata standards, and rights provenance in AI-generated and AI-assisted works. The infrastructure is being built. The question is who builds it first, whose architecture wins, and whose definitions become the industry standard.
That vocabulary race is already underway. The window to define the terms is shorter than it looks.
IV. The cap table for the song
The most useful reframe I've found for what this category of system needs to do is to stop thinking about a song as a track and start thinking about it as a cap table.
Every cap table represents a set of inputs, contributors, rights holders, restrictions, and rules about what can be done with the asset under what conditions. The cap table governs who gets paid, who gets credited, who can block a transaction, and what happens when ownership changes hands.
A pre-flight rights system has to treat the song the same way. Every source asset has a rights status. Every contributor has a claim. Every commercial pathway has permissions. Every derivative use has rules. The song is not just an audio file. It is a bundle of sound, identity, permission, monetization logic, and proof.
This is not a metaphor. It is a technical architecture. The credits, licenses, splits, AI usage logs, source asset IDs, consent records, mechanical license status, sync restrictions, audio hashes, PAD records, and audit trails that the system has to generate and maintain are not paperwork attached to a creative object. They are constitutive of it. They are what makes the creative object function as an economic asset in the markets where it needs to be traded.
For artists used to thinking about their work as art first and contract second, this framing can feel bureaucratic. That reaction is understandable and also, at this moment in the industry, a liability.
Because what the AI era has done — the thing that is genuinely new and genuinely consequential — is collapse the distance between creation and commercialization. The tools that used to take weeks now take hours. The gap between making something and releasing it has narrowed to almost nothing. In a world where that gap was six months, you had time to figure out the rights and the credits and the clearances afterward. In a world where that gap is a day, you need the infrastructure embedded in the workflow from the beginning.
That infrastructure is the work. That's the category I'm watching.
V. The unresolved tension
I want to be honest about the risk in this model, because intellectual honesty about what could go wrong is part of how you build the credibility to be taken seriously on what could go right.
A system that verifies can also exclude. A platform that scores rights risk becomes, inevitably, an arbiter of what is allowed to exist. The same gate that protects an artist from having their voice cloned without consent can also be deployed to make the platform the chokepoint for what music is commercially viable. The same audit trail that proves a small creator's ownership in a dispute can be used by an entity with more resources to bury a competitor in compliance costs.
These are not hypothetical concerns. They are the story of every major rights infrastructure build in the history of the music industry, from the ASCAP licensing model to the RIAA's aggressive enforcement posture in the file-sharing era to the platform power dynamics embedded in the current streaming royalty structure.
The people building AI rights infrastructure are not exempt from those dynamics. They are building systems that will have significant power over who participates in the commercial music ecosystem. That power needs to be held accountable in the design of the system, not patched in later when the problems surface.
What the pre-flight argument gets right, as a design premise, is the recognition that the alternative to building this infrastructure intentionally is not neutrality. The alternative is chaos: unauthorized inputs, hidden samples, undisclosed AI, royalty disputes, synthetic soundalikes, takedown cascades, and a creator class operating at enterprise velocity with hobbyist-grade protection. That is not a neutral baseline. It is a baseline that systematically advantages the actors with the most resources to fight the resulting disputes.
The goal is not a perfect system. The goal is an accountable one.
VI. Why this is a work story, not just a music story
Everything I've described about pre-flight rights infrastructure is, at one level, a music industry story. But I want to name what it actually is at a deeper level, because this is where the implications extend beyond any single vertical.
The creator economy spent a decade telling individuals to become companies. Build a brand, own your audience, monetize your identity, turn visibility into revenue. The mandate was real. The infrastructure underneath it was primitive. Creators were asked to operate at enterprise scale while using tools built for hobbyists, manage legal and financial complexity that would require teams at a startup, and build businesses in markets where the underlying asset — their creative work — existed in systems designed by and for institutional actors.
AI makes that contradiction impossible to ignore. Because if identity is labor and the self is the product — and AI can now generate plausible versions of both at scale — then authorship becomes a workplace question. Who did the work? The human? The model? The prompt writer? The original artist whose voice was absorbed into a training set? The stem owner? The publisher? The platform?
The systems we build to answer those questions will shape what creative work means as an economic category for the next generation of people who choose to do it. That is not a small stake.
The pre-flight answer is not philosophical. It is operational: document it, gate it, stamp it, route it, and make it auditable. Make the creative object explain itself.
That may be, in the end, the real future of creative work — not just making things, but making things that can account for themselves. Things that carry their provenance, their permissions, and their proof with them, legible to any system that needs to understand what they are.
The next phase of AI music will not be defined only by who can generate the catchiest hook. It will be defined by who built the infrastructure that makes the hook ownable.