Ahead of the AIM Connected conference today (September 9) including discussions on the impact of artificial intelligence on music. Here, Absolute Label Services managing director Simon Wills calls for a central, industry-owned ledger to be implemented as part of the compliance process with AI companies in order to protect artists and rights-holders…
The music industry has faced seismic shifts before, from MP3s to streaming to user-generated content. But AI-generated music is different.
This time, the machines aren’t just playing our work, they’re being trained on it, learning from it, and using it to create new tracks that can be uploaded to DSPs in minutes and, in some cases, rack up millions of streams.
The counterattack is already underway. The majors are suing AI platforms like Suno and Udio for copyright infringement. Merlin has just inked a deal with ElevenLabs, ensuring indies can participate in AI training through its Eleven Music platform.
Right now, the TDA (Training Data Agreement) framework focuses on input, giving rights-holders control over whether their works can be used for training. That’s an important start. But it’s not where the real challenge lies.
As I understand it, some of the deals on the table between music and AI companies are based around two things: an equity stake in the AI company and/or a cut of its overall AI music generation revenue.
At present, I’m not aware of any mechanism to generate additional royalties from the AI recordings themselves. That’s a major problem, because it effectively means that whoever prompts the AI to create a track keeps 100% of the ongoing royalties, while the rights-holders whose copyrighted works and recordings trained the model never see a penny of that revenue.
If we don’t address this now, we risk locking ourselves into crude, blanket arrangements that ignore the actual influence of the source material and favour those with the deepest pockets. Instead, we need a system that can trace the lineage from the AI’s training data to the final work, so that if the model produces, say, a drum & bass track, a proportionally greater share of royalties flows to the rights-holders of similar works that were used in training.
From “Was it used?” to “How was it made?”
Right now, there’s no clear, workable plan for how to track exactly how an AI-generated track was created; identify its influences, whether human or AI; and distribute royalties in a way that reflects those influences fairly.
We need a system that can trace the lineage from the AI’s training data to the final work
Absolute MD Simon Wills
Some tech companies are already trying to reverse-engineer how AI tracks are made, but that’s solving the problem backwards.
The only real solution is to capture all the data at the point of creation and store it in a central industry-owned ledger for future reference.
Under this proposal, every AI music company licensed to train on copyrighted works would be required to integrate a standardised, industry-owned compliance system into their generation process.
At the moment of creation, the system would record:
– Exact Prompt & Parameters – full generation request, including reference works.
– Daisy Chain Lineage – automatic linking back to any AI track used as a reference, and from there to the human works in its ancestry.
– Audio Fingerprint – unique ID of the generated sound.
– Composition Fingerprint – chord progressions, melodic patterns, lyrical content.
– Associated Training Data Categories – genres, styles, without exposing trade secrets.
– Mandatory Inaudible Stamp – forensic watermark embedded in the audio before it leaves the AI system, legally required for compliance.
This creates a clear, auditable trail from every AI output – human or AI, and gives the industry the raw material to decide how royalties should flow.
It would then be a requirement for all distributors to check audio against the central ledger before uploading to platforms. Meanwhile, platforms would agree to flag any matches found, attach a ledger track ID and hold back relevant royalties.
A central, neutral, not-for-profit organisation would then receive the held royalties and distribute them to identified human rights holders. Any works uploaded without the necessary checks would be flagged, triggering investigation.
The neutrality of the organisation that owns and maintains the ledger is paramount. There will, as usual, be a temptation for those with the most resources to implement mechanisms like this unilaterally and for their own benefit. Industry politics and self interest has scuppered similar projects in the past. But this will only work if it is implemented collectively in a way that works for the entire industry.
Why AI companies should support it
This isn’t about building walls. It’s about building a clear, legal, and commercially viable pathway for AI music. For AI companies, this system reduces legal risk, clarifies rights, and encourages creators to embrace their tools, because every AI track would be fully documented and compliant.
The long-term payoff
With all creation data in the ledger, we’re not locked into today’s crude blanket royalty splits.
We can refine payment models over time based on actual influence, and build rules that adapt as technology changes.
Ultimately, consent without traceability is toothless.
Without prompt logging, lineage tracking, fingerprinting, and watermarking, there will never be a fair royalty system for AI-generated music.
If we don’t act now, the AI music gold rush will look like every gold rush before it: a handful of big winners, and everyone else left panning for scraps.
