last week, of atlantic ocean has debuted a new AI detection tool that allows artists to check if their music is included in the data sets used to train the AI music generator. Created by researcher Alex Reisner, the set includes more than 21 million songs, drawn from the catalogs of big stars like Bad Bunny, Taylor Swift, Beyoncé, and smaller independent artists.
After the tool was released, multiple musicians expressed anger and concern after learning their work had appeared on the set. SZA was one of them, sharing on her Instagram Story, “I looked it up and the music AI has been trained on 238 of my songs. I’m sure there are unreleased songs as well. If you’re a musician and support this depraved shit? You’re disgusted and there’s nothing you can say to me to make this okay.” She also claimed in another account that AI companies specifically exploit black artists, writing, “I still haven’t heard an AI song by a white person… Why is it so disproportionate? We have no medical or creative protection in our legislatures. It’s easiest to steal.”
Producer Kenneth Bloom (aka Kenny Beats) specifically singled out AI music company Suno in his post. “I can’t imagine going to work every day knowing I’m stealing from countless struggling musicians,” he wrote. “I can’t imagine being proud of being paid a salary that erases an artist’s work and dreams.” On Blue Sky, teenage DJ Sabrina vented her frustrations, adding, “To all of you who thought my music sounded like AI slop, that’s cool. “Have you ever thought that it was because No was using a dataset containing 22 of my songs? It’s funny that there were no accusations that my music sounded like AI slop until these datasets started being used to generate slop.”
Not all artists who responded to this report held the same view. Producer Hudson Mohawke wrote on Instagram that the uproar surrounding the discovery of the detection tool is predicated on the state of the music industry itself. “When did the entertainment and technology industries become ‘fair’ or ‘moral’?” he wrote. “As someone who has had my full songs illegally released or sampled under other people’s names in one form or another without permission for years, I say this: Keep doing it guys. You might be able to get some of your $$ back, but you might not. Those are breaks.”
While some companies, including Google and Stability, have admitted to using the dataset to train AI models, it remains unclear which developers were actively retrieving data from the database. Three of the sets featured in atlantic oceanreports that users are directed to songs via links on YouTube and Spotify, and Reisner wrote that developers often search in automated ways, “some of which allow developers to bypass logins, advertising, and other mechanisms that could potentially capture creator revenue or subscribers.” (As Reisner points out, this violates both YouTube’s and Spotify’s terms of service.) The fourth set of data he included was based on the online Free Music Archive.
Two of the most prominent companies in the AI music space, Suno and Udio, have been sued by major labels in the past (Warner, the original plaintiff in the 2024 case, later signed a licensing agreement with Suno). Earlier this month, the American Federation of Musicians sued Universal and Warner over the use of their music by AI companies.
