AI music for business: how does it work?
With AI music, a model generates the music instead of a person writing and recording it. For a business space that matters for one reason: where there is no human author there is no rights holder, and where there is no rights holder there is nothing to collect. That saves the payments to Buma/Stemra and Sena. But it stands or falls on what the model was trained on.
Why no levies attach to it
Copyright arises with a human creation. Without a composer and without a lyricist there is no work on which Buma/Stemra could collect on anyone’s behalf.
Neighbouring rights work differently but arrive at the same place. The payment obligation under Article 7 of the Dutch Neighbouring Rights Act applies to commercially released phonograms carrying a performance by a performing artist. With music generated in real time there is no performing artist and no release, so the basis for Sena is absent.
That is not a construction and not a loophole. It is the ordinary application of two laws to a situation in which the party who normally gets paid does not exist.
Where it does run into trouble: the training data
The weak spot of AI music is not the output but the input. If a model was trained on protected work without permission, the discussion shifts to that training.
On 31 July 2026 the Regional Court of Munich I ruled largely in favour of the collecting society in GEMA against Suno, and that ruling was about exactly this. The court also held that it had jurisdiction over training that physically took place in the United States, because the output reaches European users.
The practical lesson for a business owner: ask every AI music provider what the model was trained on, and ask whether that is on paper. Anyone without an answer is selling you a risk.
What AI background music in a business is and is not
It is functional music: it gets heard, not actively listened to. Not lift music from 1994, but not something that demands attention either. For a shop, showroom, practice or waiting area that is exactly what you want, and generative AI is now good enough for it.
It is not meant for places where the choice of music is part of the identity. In a bar with a character of its own the owner picks the music, and an automatically compiled station adds nothing.
How RadioMatic does it
The music comes from a model built by a third party. The model maker states publicly that it trained exclusively on royalty-free sources, licence-free catalogues, public domain material and commissioned or licensed work.
What we can verify of that ourselves and what we cannot is set out in our legal memorandum. Including the part we cannot check — because that belongs there too.
Read the legal reasoningQuestions that come with this
Ready to hear it for yourself?
Request access to the beta. Your own station for your business, 14 days free trial, cancel monthly after that.