Suno has released v6, the company’s first music-generation model trained using licensed catalogs after direct cooperation with record labels. The training data for v6 includes licensed material from Warner Music Group, BMG and Believe. According to Suno, models that were trained on scraped internet data will be retired.
What v6 does
Suno says v6 handles musical genres that previous versions struggled with, delivering improvements in stylistic accuracy and fidelity. The company also notes an important limitation remains: v6 cannot deliberately produce a wrong or intentionally off-key note, meaning it will not reproduce errors that live performers might use deliberately.
The labels’ involvement
Notably, three record labels that previously sued AI music companies are now part of Suno’s supply chain through licensing agreements. Rather than continuing a fight over whether training the models was legal, discussions shifted to what royalty rates and licensing terms should apply.
The labels appear to have done the business calculation that banning AI music was not realistic, and that licensing provides a better revenue stream than litigation. As a result, the dispute moved from legal prohibition to negotiation over fees and revenue sharing.
Why this matters
This shift changes the industry dynamic: labels are not merely opponents of AI music but stakeholders with a financial interest in its success. The percentage of royalties they collect may become a meaningful income source, and that aligns the labels’ incentives with the growth of AI-generated music.
What comes next
The practical effects of Suno’s model transition and the new licensing deals on market practice and royalty levels will become clearer in the coming months. Retiring models trained on non-licensed data could have legal and technical consequences for actors who relied on scraped datasets.
In short, Suno v6 and the record labels’ licensing choices mark a shift in the AI music sector from litigation toward monetization through licensing and revenue sharing.
