ARTFEED — Contemporary Art Intelligence

AI Music Training Datasets Expose Massive Unlicensed Use of Copyrighted Songs

ai-technology · 2026-06-21

The Atlantic has revealed a significant investigation showing that four major song collections—12 million, 9 million, and two with over 100,000 tracks—are being used in the AI sector to train generative music models. These datasets include songs from popular artists like Bad Bunny, Nirvana, Taylor Swift, Billie Eilish, and jazz icons such as Miles Davis. Interestingly, the 1998 song "You Get What You Give" by The New Radicals is included in two of them. Three datasets provide links to songs on platforms like YouTube and Spotify, violating their terms, while the fourth, from the Free Music Archive (FMA), consists of MP3 files. Both Google and Stability AI have tapped into the FMA dataset, and over 12 musicians have filed lawsuits against AI companies for copyright issues, with some artists creating tools to counter AI models.

Key facts

  • Four datasets of songs for AI training were discovered: one with 12 million tracks, one with 9 million, and two with over 100,000 each.
  • Datasets include hits from Bad Bunny, Nirvana, Taylor Swift, Billie Eilish, Pearl Jam, Elvis Costello, Sheryl Crow, the Beatles, Miles Davis, John Zorn, and Vijay Iyer.
  • Three datasets are lists of links to YouTube or Spotify; tools bypass logins and ads to download audio.
  • Google used the Free Music Archive dataset (over 100,000 songs) to train AI models.
  • Suno has generated tracks resembling Michael Jackson's 'Thriller' and Ed Sheeran's 'Shape of You'.
  • Spotify removed 75 million AI-generated tracks; Deezer says nearly half of daily tracks are AI-generated.
  • At least 12 lawsuits have been filed by musicians and labels against AI companies for copyright infringement.
  • Benn Jordan developed a tool to poison AI models by adding inaudible noise to audio files.

Entities

Artists

Institutions

  • The Atlantic
  • Suno
  • Udio
  • Google
  • OpenAI
  • Stability AI
  • Spotify
  • Deezer
  • YouTube
  • Amazon Music
  • Free Music Archive
  • WFMU
  • Tribe of Noise
  • Sony

Locations

  • Amsterdam
  • Netherlands
  • New Jersey
  • United States

Sources