SZA and producers object to music in AI training datasets

Pitchfork reported that SZA, producer Kenneth Blume, and other musicians criticized the use of their songs in datasets associated with AI music generation. Their reactions followed The Atlantic's launch of a search tool built by researcher Alex Reisner that lets artists look for their music in datasets available to AI developers. SZA said the tool showed 238 of her songs, including possibly unreleased tracks, while Blume criticized AI music companies for building products on working musicians' output. The article notes that the datasets cover more than 21 million songs and that lawsuits against AI music companies such as Suno and Udio remain active.

Key details: Published June 21, 2026 at 17:42 UTC, The Atlantic tool searches datasets used by AI music developers, SZA said 238 of her songs appeared in the data, Pitchfork reported criticism from SZA, Kenneth Blume, and other musicians.

Why it matters: The training-data fight is moving from lawsuits and labels to individual artists checking their own catalogs, which makes the AI music backlash more visible and harder to dismiss as abstract copyright policy.