Artists React to Discovery That Their Music Trained AI Models

A detection tool built by researcher Alex Reisner and published by The Atlantic has given musicians a quick way to check whether their work appears in large datasets used to train AI music generators. The tool scans over 21 million songs drawn from four accessible sets that pull from major-label stars and independent artists alike. The reaction was swift.

SZA said she found 238 of her tracks in the data, including unreleased material. “I’m certain some unreleased,” she wrote in an Instagram story, calling support for AI music “degenerate.” On a separate account, she added that she had yet to hear a “white AI song,” stating: “Why so disproportionate? We have no protection in legislature medical or creative. The easiest to steal from.”

Producer Kenneth Blume, known as Kenny Beats, targeted the AI company Suno directly: “I can’t imagine going into work daily knowing you are stealing from countless struggling musicians.” DJ Sabrina the Teenage DJ connected past accusations that her music sounded like “AI slop” to the presence of 22 of her songs in one dataset.

Not everyone shared the outrage. Producer Hudson Mohawke posted that the anger assumes an industry that ever operated on fairness or morality. “Since when has either the entertainment or tech industry been ‘fair’ or ‘moral’?” he wrote, referencing years of his own work being illegally released or sampled.

Three of the sets contain links to YouTube and Spotify, which Reisner notes can be scraped using methods that bypass logins and creator revenue mechanisms—in violation of both platforms’ terms. While Google and Stability have confirmed using such datasets, the full list of developers scraping them remains unclear.