This isn’t some far-off, dystopian future. As detailed in The New Yorker, creators like Nick Arter, under his alias Nick Hustles, are already racking up nearly a million plays on Spotify with songs composed using AI tools. He’s part of a wave of musicians using platforms like Suno and Udio to turn text prompts into fully produced tracks, bypassing the traditional gatekeepers of A&R execs and studio engineers. It’s a democratisation of music creation, but it’s also a legal and ethical minefield. The old systems simply weren’t built for this.
What Even Are Generative Music Rights?
For a century, the music business has operated on a relatively clear premise: a human writes a song, a performer records it, and a complex but established system of rights organisations like ASCAP and BMI ensures everyone gets their slice of the pie. These bodies track public performances, sales, and streams, collecting royalties for songwriters and publishers. It’s a model built for tangible human creation.
Now, let’s introduce generative AI. What happens when the melody is conceived by an algorithm, the vocals are synthetised, and the ‘artist’ is simply a person who typed “a sad indie-pop song about lost love, in the style of Phoebe Bridgers”? Who is the author?
* Is it the user who wrote the prompt?
* Is it the company (like Suno) that built and trained the AI model?
* Is it the thousands of artists whose music was scraped from the internet to train that model in the first place, likely without their permission?
This is the central
conundrum of generative music rights. The current framework has no clean answers. It’s like trying to figure out who owns a cake baked by a magical, self-operating oven. Is it the person who tossed in the ingredients (the prompt), the company that engineered the oven (the AI model), or the farmers who grew the wheat that taught the oven what flour is (the training data)? Right now, everyone is claiming a piece of the cake, and no one has a definitive right to it.
AI Is Already Topping the Charts (Sort Of)
This isn’t just a theoretical debate. Nick Arter’s track “I’m Letting Go of the Bullshit” hit 900,000 plays. A band called Breaking Rust, another AI-assisted project, saw their song “Walk My Walk” reach three million Spotify streams. The genie is well and truly out of the bottle. Listeners, according to one study, can only distinguish AI music from human-made music about 53% of the time—effectively a coin toss. As Arter himself asked, “does it really matter if it was A.I.?”
Well, it matters a great deal to the artists whose livelihoods are at stake. The most controversial frontier is artist voice cloning. We’ve already seen AI-generated tracks that convincingly mimic the voices of Drake and The Weeknd go viral. Whilst these might be amusing novelties, they open a Pandora’s box. If anyone can create a new song in your voice, what does that do to your brand, your artistic integrity, and your income? This isn’t sampling; it’s a form of synthetic identity theft, and our legal systems are woefully unprepared to handle it.
The platforms themselves are struggling. Spotify admitted to removing a staggering 75 million “spammy tracks” in the past year, a significant portion of which were AI-generated content designed to game the streaming royalty system. They are playing a frantic game of whack-a-mole, trying to uphold quality and licensing agreements whilst being flooded by an ocean of synthetic media.
A Potential Lifeline: Stem Licensing
So, is there a way out of this mess? One of the more pragmatic solutions being discussed is the expansion of stem licensing. Traditionally, a ‘stem’ is an individual component of a track—the isolated vocal line, the drum beat, the bassline, and so on. Producers and DJs have licensed stems for years to use in remixes.
In an AI-driven world, this model could be adapted. An artist could choose to license their vocal stem—not a recording, but the very digital essence of their voice—to AI companies. In this scenario, if someone uses an AI model trained on, say, Charli XCX’s voice to generate a new melody, she (and her label) would automatically receive a micropayment or a percentage of the royalties.
This shifts the dynamic from one of potential theft to one of consensual, monetised collaboration. It gives artists control and creates a new, potentially lucrative revenue stream. Instead of fighting a losing battle against artist voice cloning, they can participate in its evolution on their own terms. It requires a robust technical framework for tracking and payment, but it’s a commercially viable path forward that protects creators.
The Wild West of Distribution Platform Policies
Right now, artists uploading AI-assisted music face a confusing and inconsistent landscape. The distribution platform policies of services like Spotify, Apple Music, and YouTube are a patchwork of reactive rules. Some distributors require explicit labelling of AI content, whilst others are still formulating their stance.
Spotify’s massive takedown of AI-generated tracks highlights the platforms’ central dilemma. On one hand, their business model thrives on an ever-expanding library of content to keep users engaged. On the other, they are under immense pressure from major labels—the industry’s power brokers—to police copyright infringement and prevent the dilution of royalty pools with low-quality, algorithmically generated noise.
For artists, this means navigating these policies is a minefield. The risk of having your music removed without warning is high if you don’t adhere to the shifting guidelines. The recommendation for now? Transparency. Clearly label any tracks that use generative AI in their creation and stay meticulously updated on the terms of service for your chosen distributor. Don’t try to sneak it past them; the algorithm will likely catch you eventually.
Can the Old Guard Adapt? ASCAP/BMI Reforms
The industry’s legacy institutions are not standing still, but they are moving with the deliberate pace of a glacier. Organisations like ASCAP and BMI are the bedrock of music publishing, and they know they must adapt or become obsolete. The discussion around ASCAP/BMI reforms is centred on a few key areas:
1. Defining ‘Authorship’: They are grappling with how to credit works with multiple human and AI inputs. Will they create new categories of ‘prompt author’ or ‘AI curator’?
2. Data Registration: They need new systems for registering synthetic works. This includes capturing metadata about which AI models were used, what prompts were given, and what training data the model was built on.
3. Distribution Models: How do you split a royalty cheque between a human songwriter and an AI company? The current splits (typically 50% to the writer, 50% to the publisher) don’t account for a non-human creator.
These are not simple tweaks; they represent a fundamental reimagining of a century-old system. Artists using generative tools need to watch these developments closely. Joining a Performing Rights Organisation (PRO) and ensuring your work is registered—even if the rules are still fuzzy—is the first step to ensuring you’re in the system when the money eventually gets allocated.
The Crucial Role of Synthetic Media Attribution
Underpinning all of this is the concept of synthetic media attribution. In simple terms: we need a reliable way to label and trace AI-generated content. Without a clear “Made By AI” tag that is difficult to forge, the entire system falls apart. It’s about creating a chain of custody for digital creation.
Imagine a future where every AI-generated song has embedded, non-editable metadata that details:
– The AI model and version used.
– The original human prompter.
– The primary training data sources (and whether they were licensed).
– The percentage of the work that is AI-generated versus human-performed.
This isn’t just for royalty purposes; it’s for cultural and ethical reasons too. It allows listeners to make informed choices, helps platforms filter content effectively, and ensures that human artists whose work formed the basis of the training data can be identified and, hopefully, compensated. Technologies like blockchain are being explored for this, offering the promise of an immutable public ledger for creative works.
The path forward for music is not to ban AI, but to integrate it ethically. The tools are too powerful and the creative potential too vast to ignore. But this integration can’t happen in a lawless vacuum. We are in the messy, chaotic, and incredibly important process of building the legal and economic architecture for the next century of music. The artists who thrive will be the ones who don’t just use the new tools but also understand the new rules of generative music rights.
What do you think? Should an AI be able to be listed as a credited songwriter? And who should get paid when a viral hit was born from a one-sentence prompt? The floor is yours.
