Artificial intelligence
The elephant in the room
We don't have a policy on AI music. We have a structure that means you never get handed it without asking — and the reason we can't police it is the same reason nobody can police you.
The short version
An absence, not a policy
The worry isn't really that AI music exists. It's being handed it without knowing — slipped between two human artists on a service you pay for.
That requires a delivery surface. Something that picks the next track for you and has a reason to prefer one over another. Station doesn't have one.
So the answer isn't a rule we enforce, it's a thing we never built. Anyone can pay to generate AI music and host it on a server they own. Anyone can pay to listen to it. Neither is ours to police — and neither reaches you unless you go looking.
Where music can arrive from
Five ways in on one side, two on the other — and both of the remaining two require a person to have acted first.
What we don't do
Four things we deliberately skipped
- No ban
- There is no rule against hosting AI-generated music on a Station server, because there is no mechanism by which we could enforce one. Your server is yours.
- No detector
- We do not run classifiers over uploaded audio. Detection is an arms race against tools that improve monthly, and the side doing the detecting is always a step behind.
- No labels
- No mandatory disclosure field, no badge, no "AI-assisted" checkbox. A field like that is only as honest as whoever fills it in, and it invites the wrong argument about where assistance ends.
- No appeals process
- Which follows from the rest. Nothing is being judged, so there is nothing to appeal, and no queue for a human artist to sit in while somebody decides whether they are real.
The false positive is the part people underrate. A detector that wrongly flags a human musician does more damage than one that misses AI entirely — it takes someone who made something and tells them, with the authority of a system, that they didn't. Every platform that has tried this at scale has produced that outcome, and there is no appeals queue long enough to fix it.
The real reason
The rule that spares them is the one that protects you
Suppose we did want to remove AI music from the network. What would that take?
It would take the ability to reach into a server we don't own and delete files we didn't put there. A switch that takes somebody's catalog offline because we decided it should be offline.
That switch doesn't exist. Not because we chose restraint, but because the architecture never gave anyone one — your music sits on your machine, your catalog is served by your node, and the network only ever learns where to find you.
Where it lands
On the listener, on purpose
Somebody has to decide what's worth hearing. On a platform that's a model, tuned to objectives you can't see and wouldn't be told about. Here it's you, and the people whose taste you already trust — the critic, the DJ, the blog, the friend who sends links.
That's more work. It's also the thing that was taken, and the reason the space where those people operate was left empty rather than filled in with a better algorithm.