Spotify has released a one-year update on its pledge to keep low-quality AI-generated music off its platform. The report was published on October 20, 2026.
The streaming company said it excluded more than 50 million slop and spam tracks from its recommendations this year. It also removed 75 million phony tracks outright.
AI tools have made it easier to produce songs in large numbers. Some people have used those tools to cheat streaming services.
A Prison Sentence for Streaming Fraud
A North Carolina man named Smith was recently sentenced to 18 months in prison for a scheme tied to AI music. According to the U.S. Department of Justice, he created hundreds of thousands of AI-generated songs.
Smith then used bots to play those songs billions of times. Over the course of the scam, he collected more than $8 million in royalties.
He released the tracks under made-up artist names. These included Calm Knuckles, Calorie Screams, Calypso Xored and Calliope Erratum.
Another project drew attention on Spotify under the name Velvet Sundown. Its streaming biography described “shimmering tremolos, warm tape reverbs, and the gentle swirl of organs.”
The project later admitted the music was made with AI. YouTube commenters responded with jokes about the band’s machine origins.
AI Music Projects Built With Musicians
Not every use of AI in music has been aimed at gaming streaming platforms. In the 1980s, trombonist and composer George Lewis created Voyager, a computer-driven improvising orchestra.
Voyager could respond to live musicians in real time. It could also head off in its own musical directions.
A newer example comes from Dadabots, a team of two music hackers. They trained a neural network on extreme metal music.
Their project, called “Relentless Doppelganger,” has streamed technical death metal online 24 hours a day since 2019. It runs on SampleRNN, a recurrent neural network first built for speech, which creates audio one tiny slice at a time.
To sort through the large amount of output, the team built a tool called D.O.M.E., short for Disproportionately-Oversized Music Explorer. It uses machine learning to group and display the results so the best parts can be found.
At the other end of the scale is Forgotify, a website that surfaces Spotify tracks that have rarely or never been played. A sample of about 25 tracks turned up obscure metal, vintage jazz and blues, and many songs with “drone” in the title.
One find was Shorty Rogers’ “Cool Sunshine,” from the 1955 album East Coast-West Coast Scene. Another was a sound effect called “Very Distant City Traffic Drone with Sparse Metal Movement.”
Spotify’s report puts its total action this year at more than 125 million tracks. That figure counts the 50 million kept out of recommendations and the 75 million removed from the platform.
