Music combines everything that makes a field vulnerable: a fully digital product, platform-based distribution, already fragile remuneration, and automatic generation that reached a convincing level faster than in most other arts. It's also a sector where AI serves as a production tool for a huge number of creators. Both realities coexist.
We covered the legal side in our article on Suno, Udio and the Sacem. Here's the practical side: what it actually changes for someone making music.
What AI really brings to creators
Let's start here, because it's massive and often glossed over in discussions on the topic.
Studio-free production. Stem separation on an existing recording, automatic mastering, cleaning up takes, replacing instruments. These operations used to require a sound engineer and a studio. They're now doable from a bedroom.
Rapid prototyping. Generating a demo to test an arrangement idea before actually recording it. AI serves as a draft, not a finished product.
Filling skill gaps. A songwriter who writes well but plays no instrument can produce a backing track for their lyrics. A producer without a singer can lay down a guide vocal.
For an independent artist with no budget, these tools remove barriers that had nothing to do with talent. It's the same phenomenon described in our article on video generation: the number of intermediaries between an idea and its final form shrinks.
What threatens, concretely
Three distinct threats, which need separating because they don't call for the same responses.
Functional music. This is the most immediate and best documented. Hold music, video background tracks, advertising jingles, venue soundscapes. This market paid a large number of musicians without fame, and it's exactly the segment where a generated track suffices. The criterion isn't artistic, it's functional: the music has to fill a space without disturbing.
Platform saturation. When producing a track costs nothing, the catalogue explodes. The consequence isn't that listeners accidentally hear generated music; it's that discoverability collapses for everyone. An artist drowned among millions of tracks isn't competed against, they're invisible.
Voice cloning. This is the most sensitive issue, because it touches identity. The voice is the most personal thing an artist has, and reproducing it without permission raises a question that goes beyond classic copyright.
Three things, and they show where value is shifting. Live performance: a concert can't be generated, and its share of artists' income has never been more decisive. The person: you follow an artist for what they say, their journey, their way of being in the world. That's exactly what we noted in our article on art in the age of AI. The community: a direct relationship with a supportive audience can't be substituted.
What platforms are starting to do
The sector is organising. Some streaming services have rolled out detection tools that identify tracks produced by the most widespread generators, and some flag them explicitly. Other platforms have banned fully generated music or vocal identity theft.
Deals have also been struck between generators and major labels, as we've documented. The direction is toward a licensing system, with the usual caveats: these agreements primarily cover major catalogues, and the share that flows down to individual artists remains a point of contention.
What to take away
Music is showing in fast-forward what awaits other creative fields. Value is shifting from production toward what can't be produced automatically: presence, relationship, identifiable intent.
It's a painful redistribution, because it hits first those who made a living from honest, unglamorous production work. It relatively favours those who already have an audience, which isn't a very fair mechanism.
What remains true, and what tools don't change: no one has ever looped a track because it was technically well produced. You come back to it because it says something you needed to hear. That part isn't threatened, but it can't be delegated either.