Type a sentence, wait a few seconds, receive a song. Text-to-music systems have crossed the threshold from novelty to product, and the music industry is now living through its most consequential copyright argument since file sharing. The question sounds philosophical but is intensely practical: who owns a song that no human wrote — and who gets paid when it plays?
The authorship problem
Copyright, in most of the world, protects works of human authorship. The U.S. Copyright Office has repeatedly taken the position that material generated entirely by a machine, with no meaningful human creative control, is not copyrightable at all — a stance echoed in court decisions on AI-generated works. The EU's framework points the same direction: no human author, no author's right.
That leads to a strange place. A fully AI-generated track may belong to no one — it can drop into a legal grey zone closer to the public domain than to a record label's vault. Prompting alone ("make me a sad synthwave song") is unlikely to count as authorship; substantial human editing, arrangement or performance might. The line is being drawn case by case, and it will take years to settle.
The training-data fight
Ownership of the output is only half the war. The other half is the input: today's music models learned from enormous catalogues of existing recordings. In 2024 the major labels, through the RIAA, sued leading AI music generators, alleging their models were trained on copyrighted recordings at massive scale without licence. The defendants argue fair use; the plaintiffs call it industrial-scale copying. Parallel fights are running in Europe, and some jurisdictions are moving toward explicit text-and-data-mining rules with opt-outs.
However these cases end, one outcome is already visible: provenance is becoming a feature. Labels, publishers and broadcasters increasingly need to know — and prove — where a recording came from, what it was trained on, and who holds the rights.
Why this matters if you play music in a business
For a venue, the appeal of AI music is obvious: cheap, endless, royalty-free background sound. The risks are less obvious:
- "Royalty-free" is a claim, not a fact. If a generator was trained on unlicensed recordings and a court later disagrees with its fair-use theory, catalogues built on it inherit that uncertainty.
- Soundalikes and voice cloning carry their own liability. A track that imitates a real artist's voice or style can trigger publicity-rights and unfair-competition claims even where copyright doesn't reach.
- No author also means no accountability. With licensed music there is a chain — writer, performer, label, broadcaster — that stands behind what plays in your room. With anonymous AI output, that chain ends at a prompt.
- Your atmosphere is your brand. Even setting law aside, model-generated filler is optimised to be average. Rooms notice.
The cheapest music in the world is expensive if a rights holder disagrees with how it was made.
Where SpinLoud stands
SpinLoud's position is simple: music in a commercial space should have a traceable human chain behind it. The channels on our platform are live radio programmed by music directors and produced by independent, licensed broadcasters — real recordings, real rights holders, real accountability. We watch the AI-music cases closely (some of the tooling is genuinely exciting, especially for production and audio processing), but we will not put unattributable, unlicensed model output on air in your venue.
AI will absolutely be part of music's future — as an instrument in human hands. The courts are now deciding what happens when the hands are removed. Until they do, the safest and, frankly, the best-sounding choice for a business is music with an author.
SpinLoud carries 210+ licensed radio channels programmed by human music directors. Start your free 7-day trial and hear the difference an author makes.