Artificial intelligence is forcing the music industry to answer an old question in an entirely new environment:
What is a song actually worth?
Not the recording. Not the singer’s image. Not the platform carrying it. The song itself.
The melody somebody conceived. The lyric somebody struggled over. The harmony, structure and emotional architecture that existed before there was a master recording to stream, remix, imitate or feed into a machine.
That distinction has become increasingly important because AI is no longer standing outside the music business asking to be admitted. It is already inside.
The Song Is Not the Leftovers
On 30 September 2026, independent music publishing organisations IMPF and IMPEL published a joint framework for generative-AI licensing. At its centre is a deceptively simple proposition: the song must be properly valued.
The organisations argue that, unless there are compelling reasons otherwise, AI training and exploitation should value the composition at least equally with the recording. Their framework also calls for transparent deductions, clearly defined licence scopes, attribution mechanisms and appropriate royalties from downstream uses of AI-generated music.
That matters because music has historically developed different economic pipelines for recordings and compositions. An AI system, however, can encounter both.
It can encounter the sound recording. But it can also encounter the musical information embedded within it: melody, harmony, lyric, arrangement, structure and rhythm.
So if a machine derives commercial capability from human-created music, we have to ask where the resulting value travels.
And more importantly: does it ever travel back to the person who wrote the song?
AI Is Not the Enemy. Invisible Extraction Is.
This distinction matters to me because I do not approach artificial intelligence as an enemy of creativity.
I use emerging technology. I am interested in hybrid production. I believe artists can use new tools without surrendering the human authorship at the centre of their work.
AI-assisted creativity is not the same thing as replacing the creator.
A musician can use technology to realise an idea. A songwriter can experiment with production. A producer can manipulate sound. Artists have always adopted new instruments.
The synthesiser did not abolish composition. Digital recording did not abolish musicianship. A camera did not abolish the photographer.
The question is not whether technology participates. The question is who is directing the creative act, whose intellectual property is being used, who consented, and who receives the value created from it.
Consent Without Accounting Is Not Enough
We therefore need to move beyond the simplistic argument of “AI versus artists.” The real conversation is considerably more sophisticated.
Did the creator authorise the use? Can the work be identified? Can its influence be traced? Is the licence narrow enough that permission for one purpose does not quietly become permission for ten others? What happens when an AI-generated output earns money? And who audits the calculation?
These are not anti-technology questions. They are ordinary business questions.
If a hotel uses music, licensing exists. If a broadcaster uses music, licensing exists. If a film synchronises music, licensing exists. If digital services stream music, licensing exists.
The arrival of a technologically complicated customer should not suddenly make the principle of paying creators technologically inconvenient.
The Independent Artist Has the Most to Lose
Large catalogues have lawyers, licensing departments, metadata systems and negotiating leverage. Independent creators frequently have themselves.
That creates an uncomfortable imbalance. The people whose work may contribute to enormously valuable technologies can be the people least equipped to discover that their work was used, negotiate the terms or audit what they are owed.
That makes accurate metadata, attribution and rights administration even more important.
Because in an AI economy, a song that cannot be reliably identified can become a song that cannot be reliably paid.
Technology Should Make Royalty Accounting Better, Not Worse
There is an irony here. We are discussing some of the most sophisticated computational technology humanity has ever created while simultaneously entertaining the possibility that identifying and compensating the human beings whose work helped make that technology useful is somehow too complicated.
Surely artificial intelligence should allow us to build better attribution, not less.
Better metadata. Better matching. Better royalty accounting. Better detection of unauthorised imitation. Better systems for distinguishing human-created, AI-assisted and substantially AI-generated work.
The technology capable of analysing millions of recordings should also be capable of helping us identify whose music contributed value.
The obstacle cannot permanently be technical capability. Eventually, it becomes a question of commercial will.
Innovation and Rights Are Not Opposites
The music industry does not have to choose between technological progress and creators. That is a false choice.
Innovation can flourish while creators retain meaningful rights. AI companies can build extraordinary tools while licensing intellectual property. Artists can experiment with technology without surrendering authorship. Fans can experience entirely new forms of music without being deceived about who or what created them.
And songwriters can participate economically in technologies partly built upon the accumulated language of human music.
The future does not have to be human or machine. It can be human with machine.
But partnership has a condition.
Everybody at the table must be recognised.
And if the song helped build the future, the songwriter should not be the last person paid for it.
Aremuorin
Thoughts. Culture. Truth.
Further reading: IMPF and IMPEL, “Fair Licensing for Generative AI,” published 30 September 2026.
Leave a comment