Last week, the Atlantic debuted a new AI detection tool that allows artists to see, with a brief name search, if their music appears in data sets used to train AI music generators. Created by researcher Alex Reisner, the tool draws on four different sets available to AI developers. These sets, which Reisner emphasizes is not comprehensive, encompasses over 21 million songs. They pull from the catalogs of massive stars like Bad Bunny, Taylor Swift, and Beyoncé, and smaller, independent artists alike.
In the days following the tool’s launch, multiple musicians have expressed anger and concern over discovering their work in the data sets. SZA was among them, sharing in an Instagram story: “Jus checked and music AI has trained off 238 of my songs. I’m certain some unreleased. If your a musician and you support this degenerate shit? Your disgusting and there’s NOTHING YOU COULD EVER SAY TO ME TO MAKE THIS OKAY.” On a separate account, she also asserted that AI companies specifically exploit black artists, writing: “I AINT HEARD A WHITE AI SONG YET.. why so disproportionate? We have no protection in legislature medical or creative. The easiest to steal from .”
In his own post, the producer Kenneth Blume (FKA Kenny Beats) specifically called out the AI music company Suno. “I can’t imagine going into work daily knowing you are stealing from countless struggling musicians,” he wrote. “I can’t imagine being proud to earn a paycheck obliterating the work and dreams of artists.” On Bluesky, DJ Sabrina the Teenage DJ aired her own frustrations, writing: “To everyone who thought my music sounded like AI slop, did you ever think it was because Suno was using a dataset that contained 22 of my songs? It’s funny how there were no accusations of my music sounding like AI slop until these datasets started getting used to generate slop.”
Although some companies, including Google and Stability, have admitted to using these data sets to train their AI models, it remains unclear which other developers have actively taken from the databases. Three of the sets featured in the Atlantic’s report led users to songs via YouTube and Spotify links, which Reisner wrote that developers often search with automated methods, “some of which allow developers to bypass logins, advertisements and mechanisms that might earn money or subscribers for creators.” (That, Reisner pointed out, would be a violation of both YouTube and Spotify’s terms of service.) The fourth data set he included was based on the online Free Music Archive.
Not every artist who responded to the report shared the same perspective. On his Instagram story, the producer Hudson Mohawke wrote that the uproar around the detection tool’s findings presupposes the very conditions of the music industry. “Since when has either the entertainment or tech industry been ‘fair’ or ‘moral’?” he wrote. “I say that as someone who has had my own full songs illegally released or sampled under other ppls names for years in some form or another without clearance, just ppl going ahead n doing it. Maybe you can scrape back some of the $$ maybe you can’t. Them’s the breaks.”
Suno and Udio, two of the most prominent companies in the AI music space, have both been sued by a cohort of major labels in the past. Earlier this month, the American Federation of Musicians also sued Universal Music Group (UMG) and Warner Music Group (WMG) over the use of their music by AI companies. Warner signed a licensing deal with Suno last year.
