“Create never before heard music in 1-click using AI,” reads MusicFlow’s slogan. There’s a problem, though: this does sound like something you’ve heard before, because it’s been trained using millions of songs made by real artists, including Milwaukee artists.
MusicFlow is just one of many generative AI platforms that lets users create “original” songs by typing in a prompt. This isn’t the only one that exists—there’s Suno, Udio, OpenAI, along with many others.
Alex Reisner, staff writer at The Atlantic, discovered four different datasets filled with songs that are being used by these AI trainers, without the knowledge from the artists, record labels, and streaming platforms involved, if any.
One of these datasets contains 12 million songs, the other one has 9 million, and the other two have a couple hundred thousand. Reisner wrote that “the 12 million track dataset would take 91 years to listen to on its own.”
In the Atlantic article, there’s a free search bar that lets you look up your own name to see if your music has been used on any one of these datasets. It is unclear from here who exactly has used these downloadable sets, and what they have been used for.
Carl Nichols, Milwaukee-based blues artist recently made a post on Facebook after his artist name Buffalo Nichols turned up 13 search results:
Today I learned the Atlantic shared a searchable list of songs being used to train models for generative AI music (Suno and Udio). I also learned 13 of my recordings are in that list. I’ve talked a lot about how most of the blues music being PURCHASED on iTunes is AI generated (every song in that list in slide 2 is AI). A few people stand to make a lot of money from AI and most of us will suffer. I know some people are doing the work to go after the bad actors but I really want us as a culture to start shaming the losers who make and listen to this garbage. If you can’t educate them, start bullying these people.
The Atlantic article also mentions that some of these AI developers claim they only use music downloadable on free websites. But Reisner found that three of the sets are actually taken straight from Spotify or YouTube—really anywhere that music is uploaded for “free.”
“AI developers download the actual audio using tools that automate the job, some of which allow developers to bypass logins, advertisements, and mechanisms that might earn money or subscribers for creators,” Reisner wrote. “Such tools violate the terms of service of these platforms. (The fourth dataset, the Free Music Archive collection, is distributed with MP3s.)”
On any one of these datasets you can find music by both major label artists and independent musicians, from all types of genres and decades.
Sony Music Entertainment, Universal Music Group (UMG), and Warner Music Group (WMG) filed a lawsuit against AI tech companies Suno and Udio for copyright infringement in 2024. In 2025, UMG and WMG settled.
On June 5, the American Federation of Musicians sued UMG and WGM for licensing their work to Suno and Udio, arguing that while UMG and WMG settled and were compensated, the individual artists were not.
Hagens Berman, a law firm famous for taking on the Big Tobacco case, is now exclusively representing independent artists through ongoing class-action settlements.
This establishes a strange competition between artists and their own selves, between the music they actually made and the AI songs that sound like it. With artists’ intellectual and creative property at risk, this affects their revenue, and, on a greater and much scarier scale, individuality and the protection behind art itself.
When Milwaukee pop/soul and R&B artist Micah Emrich searched for his own name, 40 of his songs were included in the list.
“I wasn’t shocked since I’d seen other artists speaking out about their songs being listed in the database.” Emrich said, “but what did shock me was seeing how many of my songs were being used. Forty of my songs were listed…some of them aren’t even on DSPs (Digital Service Provider) anymore, so it’s confusing as to how they got access to them. I actually said out loud ‘HOW!?’ I never gave consent for my music to be used to train AI, and am definitely not getting any sort of compensation for this. It feels criminal. Not just for me, but all artists who haven’t given clear consent for their IP to be used in such a way.”
Again, both represented and independent artists are facing the same exact battle.
“I don’t think it makes a difference how big of an artist you are. If you aren’t consenting to your music being used, then it’s wrong. Point, blank, period,” Emrich said. “I saw SZA recently spoke out against it when she found that over 200 of her songs were being used to train AI. Everybody’s at risk.”
“Jus checked and music AI has trained off 238 of my songs. I’m certain some unreleased,” reads the statement that SZA posted on her Instagram story. “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.”
If you’re an independent artist looking for a way to protect your music, class-action lawsuits are still currently available for sign-up via Hagens Berman.
Want more Milwaukee Record? Subscribe to our free weekly newsletter and/or support us on Patreon.