Q&A with Grimes, who invited anyone to make Grimes songs using AI tools, on “open-sourcing” her voice, her obsession with AI, and five GrimesAI-created tracks
the producer and pop singer who has long been enthralled with visions of the future — sees opportunity. https://www.nytimes.com/... @cfarivar : “We were going through my old college sketch pads last year and we found a bunch of A.I. theory.” Wait what. https://www.nytimes.com/... Joe Coscarelli / @joecoscarelli : GRIMES REVIEWS A.I. GRIMES SONGS https://www.nytimes.com/... [image]
Context & Ripple Effects
A month after offering fans a 50/50 royalty split on any AI song made with her voice, Grimes is following through in public: this NYT interview has her reviewing five finished GrimesAI tracks and talking through the college-era A.I. theory sketch pads behind the experiment. The move runs against the grain of an industry where labels are still defining their stance — UMG's Lucian Grainge is separately positioning generative AI as the next revenue engine after streaming (per the New Yorker profile).
The technical groundwork predates the moment: OpenAI's Jukebox neural net was generating rudimentary artist-style vocals back in 2020, and by the time of Grimes' offer, text-prompt music tools were close enough that a solo artist could plausibly license herself instead of waiting for label infrastructure. Her August follow-up Q&A on the "c" persona shows the experiment becoming an ongoing identity project, not a one-off stunt.
First-order effects
- Anyone can now produce and release Grimes-branded songs with her blessing, with revenue split evenly between creator and artist — the five tracks she reviews are the first proof the pipeline works end-to-end.
- Grimes converts her own voice from a controlled asset into an open input, trading exclusivity for volume of derivative work she co-monetizes.
Second-order effects
- Labels watching the experiment get a working template for licensed-voice deals — Grainge's push to monetize generative AI at UMG now has an artist-side counterpoint showing consent-plus-revenue-share as an alternative to litigation over unauthorized voice clones.
- Synthetic-voice and text-to-song toolmakers gain a high-profile legitimacy case: a real artist treating their models as distribution rather than threat.
Third-order effects
- If the split model proves viable, artist voice becomes licensable IP with formal royalty terms — shifting the industry fight from whether AI vocals are allowed to who signs the contract and at what percentage.
- The pattern points toward a two-tier market: major labels negotiating catalog-scale voice licenses while independent artists self-license directly through platforms.
The trend: Artists are moving faster than labels to define the terms of AI voice use, turning their own identities into the first licensed datasets of the generative-music economy.