The grueling fight over who profits from AI music

When Machines Write Songs, Who Collects the Royalty?

Ecorescuezone.com – The streaming economy that has defined music consumption for two decades is now being stress-tested by a technology that can compose, sing, and produce a full track in seconds. Generative AI tools have moved from novelty to commercial scale, and the question they force on the industry is no longer hypothetical: if a machine learned its craft from millions of human recordings, who is owed payment when that machine’s output earns revenue?

For roughly a hundred years, the answer was straightforward. The music business ran on a permission-and-payment framework. Sampling a drum break required clearing the original composition and master. Placing a track in a film, a commercial, or a television episode triggered a chain of licensing fees flowing to songwriters, publishers, and record labels. An entire infrastructure of rights management, royalty accounting, and collective administration grew up around the principle that creative labor commands compensation.

That infrastructure is now under siege.

The Platforms That Changed the Conversation

Services such as Suno and Udio allow a user to type a short text description and receive back a finished song — complete with vocal performance, lyrical content, and full instrumental arrangement. The user base spans casual experimenters, social-media content producers, independent artists, and Grammy-winning producers. The scale is no longer marginal. In February, Mikey Shulman, co-founder and chief executive of Suno, announced on X that the company had crossed two million paying subscribers.

Two million subscribers is not a laboratory curiosity. It is a revenue stream large enough to reshape market expectations, and it signals that AI-generated audio has crossed from the fringe into the mainstream of music production.

The Transparency Gap

Before any compensation question can be answered, a more basic one must be settled: what exactly did these models learn from? For much of the past several years, the training corpora behind major music-generation systems have been opaque. Artists, scholars, and litigants have struggled to determine which recordings fed into which architectures, making it difficult to construct either a fair-use defense or a copyright-infringement claim.

Alex Reisner, a journalist at The Atlantic, has spent considerable effort closing that gap through his AI Watchdog project, a suite of search tools that lets musicians scan published training datasets and locate their own recordings within them. Artists including SZA have used the tool to confirm that their work appeared in the corpora behind commercial AI systems.

“I think it’s really difficult to discuss the potential of the technology, the risks of the technology, without having more information out there about how it’s trained. And so my purpose with AI Watchdog and these search tools is just to try to give people access to… some of the raw data.”

Reisner notes that many performers remain genuinely startled to discover their recordings were ingested without their knowledge or consent.

“Which I think really reflects the strength and persuasiveness of the narratives that these tech companies are putting out there about how they’re just creating this kind of magical resource when really what they’re doing is taking everyone’s stuff and kind of reorganizing it, remixing it.”

The opacity is not merely an academic concern. In court filings, Suno has acknowledged that its training data comprised “essentially all music files of reasonable quality that are accessible on the open Internet,” supplemented by other available sources. Whether that sweeping ingestion qualifies as fair use remains contested, but even granting the fair-use characterization, the compensation question does not dissolve.

Dilution in the Streaming Pool

Even if no individual song is copied, the economics of streaming distribution create a second, subtler form of harm. Streaming services collect subscription fees from listeners and then allocate a portion of that pooled revenue across every play of every track in their catalogs. The more tracks compete for plays, the thinner each individual share becomes.

Krystle Delgado, an entertainment lawyer, independent artist, and podcaster who tracks the legal and financial dimensions of the AI-music dispute, frames the problem in blunt arithmetic terms:

“This matters because it’s actually the same pool. Everyone’s music, the AI music, the human-made music, goes into the same place, and then it’s divided up.”

Tens of thousands of AI-generated tracks are now appearing on major streaming platforms. Delgado argues that the sheer volume of machine-produced audio, competing for the same finite pool of listener attention and subscription dollars, will progressively dilute the per-play payout available to human composers and performers.

“And so right now we just have a numbers issue. And so is it taking away from the other music that is already there? Yes. And just by simple math, at some point the AI music is just gonna drown out the human-made music.”

The Compensation Question That Has No Precedent

The deeper structural issue is one the industry has never had to confront. If an AI model trained on the entire recorded-music canon generates a new song, and that song is streamed millions of times, where was the economic value actually created? How much of it belongs to the original rights holders whose recordings shaped the model? How much to the company that built and operates the system? How much to the user who typed the prompt?

No existing royalty framework — whether the mechanical-reproduction model, the performance-rights model, or the streaming-distribution model — was designed to answer that question. The industry’s century-old architecture of identifying ownership, attributing contribution, and routing compensation simply has no slot for a machine that learned from everything and produced something new.

The legal battles now unfolding in federal courts, combined with the rapid commercialization of tools like Suno and Udio, mean that the answer will be written not by legislators or industry committees but by judges, arbitrators, and the market itself. Whatever shape that answer takes, it will determine whether the next generation of music economics rewards the humans who built the canon or the systems that learned from it. The stakes are not merely financial; they define whether creative labor retains economic meaning in an age when a text prompt can produce a finished song.