How Musicians Can Get Paid for Training AI
TL;DR
Music royalties usually follow use: sales, streams, radio, covers and karaoke all have rules. AI training breaks that logic because a song may be used once in training but keep shaping model outputs afterward. Sureel, now acquired by Warner Music Group, is working with STIM on training-data rules: free use, limited influence or opt-out, with licensing fees tied to how material is used in training.
Nauti's Take
Attribution is being pitched as the clean technical fix for music and AI. That is too convenient.
Without open rules, independent audits and simple participation models, artist payment becomes another rights-management maze that creators cannot inspect. The promising path is smaller, clearly licensed models where artist collectives share real upside, not private deals that mainly protect catalogs and AI companies.
Briefingshow
This is bigger than a new royalty mechanism for musicians. The music industry is testing whether AI companies can treat culture as raw input or whether creative work keeps a claim on the value it creates inside models. If the rules are too simplistic or opaque, they will lock in platform power instead of strengthening creators.
Sources