How Musicians Can Get Paid for Training AI
TL;DR
Music rights already handle many kinds of use: sales, streams, radio play, cover versions and karaoke all have payment logic behind them. Generative AI complicates that model because the use may happen once during training, or continue every time the model produces music shaped by that data. Sureel and SoundVerse are working on attribution and payment systems so training music is not treated as a one-off data grab.
Nauti's Take
Music has a better starting point than many creative fields because the industry already knows how to turn tiny uses into payments. That makes training less like a technical preprocessing step and more like a new category of monetizable use.
But without transparent measurement, audit rights and clear payout rules, this can easily become a convenient fig leaf for AI companies.
Briefingshow
The issue is not only copyright, but the economic foundation of creative work. If AI models learn from music and keep generating outputs shaped by that training, a one-time license may not be enough. The real test is whether attribution can become technically reliable enough to support recurring payments.
Sources