SZA, Kenneth Blume Denounce AI Training on 21 Million Songs as Tool Flags 238 SZA Tracks
Updated
Updated · Pitchfork · Jun 21
SZA, Kenneth Blume Denounce AI Training on 21 Million Songs as Tool Flags 238 SZA Tracks
2 articles · Updated · Pitchfork · Jun 21
Summary
A new Atlantic-backed search tool built from four music-training datasets let artists check whether their work was used by AI developers, prompting SZA to say 238 of her songs were included, possibly including unreleased material.
Kenneth Blume, formerly Kenny Beats, used the findings to accuse Suno of stealing from struggling musicians, while DJ Sabrina the Teenage DJ said 22 of her songs appeared in the datasets.
The database is not comprehensive, but it spans more than 21 million songs; researcher Alex Reisner said three of the sets point to YouTube and Spotify links that developers often scrape with automated methods that can violate platform terms.
Google and Stability have acknowledged using such datasets, though it remains unclear which other developers did; Hudson Mohawke argued the backlash reflects long-standing unfairness in music and tech.
The disclosures add to mounting legal pressure on AI music firms: Suno and Udio have already been sued by major labels, and the American Federation of Musicians recently sued Universal and Warner over AI-related music use.