How Musicians Can Get Paid for Training AI
TL;DR
Music rights already price many kinds of use: sales, streaming, radio, covers and karaoke. Generative AI breaks that logic because training may happen once, while the work’s influence can shape every later output. Sureel, now acquired by Warner Music Group, is working with Swedish collecting society STIM on reports that could track training permissions, influence limits and licensing fees.
Nauti's Take
The direction is right: one-off buyouts look like cheap absolution for music AI. But attribution must not become another opaque machine that artists only understand after the money has already moved.
The real test is not whether the dashboard looks clever, but whether independent reviewers can see why one work gets paid and another does not. Without that, fairness becomes PR with a formula.
Briefingshow
This is not just about retroactive compensation; it is about defining what creative use means in the AI era. If training is treated as a one-time data import, AI companies keep most of the value. If influence is paid per output, the industry gets a new royalty layer that could be fairer, but only if it is auditable and not controlled by private black-box deals.
Sources