The files, obtained by 404 Media from a hacker identified as “ellie.191,” detail specific scraping operations conducted between 2023 and 2024. Documentation reveals that Suno utilized the third-party service Bright Data to extract audio from YouTube and intentionally sought out a cappella versions to isolate vocal tracks. One internal file explicitly cites a collection of 2,013,545 YouTube Music clips used for model training.
Leaked Files Reveal Extent of Suno's Music Scraping
A massive data breach at AI music generator Suno has confirmed the company scraped millions of copyrighted tracks from platforms including YouTube Music, Genius, and Deezer. The leaked source code and instructions provide concrete evidence for ongoing legal battles regarding whether the startup’s training methods constitute fair use or theft.
These revelations directly challenge Suno’s previous stance in litigation brought by the Recording Industry Association of America. While the company has argued that its use of public internet data falls under fair use doctrine, the RIAA maintains that the startup illegally circumvented digital copyright protections. Beyond YouTube, the leaked scraping instructions implicate a wide array of sources, including Pond5, Jamendo, Freesound, and the International Music Score Library Project.
Comments (0)
No comments yet. Be the first!