Updated
Updated · The Verge · Jul 15
Hacked Suno Files Reveal 2 Million YouTube Music Clips in AI Training
Updated
Updated · The Verge · Jul 15

Hacked Suno Files Reveal 2 Million YouTube Music Clips in AI Training

3 articles · Updated · The Verge · Jul 15

Summary

  • 2,013,545 YouTube Music clips appeared in leaked Suno files, which also showed the AI music generator trained on millions of scraped songs and lyrics from Deezer, Genius and other platforms.
  • 2023 and 2024 source code and scraping instructions detailed pulls from YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound and IMSLP, with code also seeking a cappella tracks and roughly 1 million hours of podcasts.
  • The leak bolsters record-label claims that Suno used copyrighted works and may have bypassed YouTube protections; Suno has argued in court that training on publicly available internet music is fair use.
  • Customer email addresses, phone numbers and Stripe-related payment details were also exposed, though Suno said the November 2025 incident was quickly contained and did not warrant individual breach notifications.

Insights

A hack exposed Suno's training data from YouTube. Are other AI giants hiding similar copyright risks?
Suno is valued at $5.4B but faces $9B in damages. Is this tech's biggest legal gamble yet?