Source Code Hack Reveals Suno's AI Was Trained on Millions of YouTube Songs

A source code hack reveals that Suno AI trained its music generator by scraping millions of songs and lyrics from YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound, and podcast RSS feeds. This data confirms suspicions from record labels and artists regarding the use of copyrighted material. While the record industry has filed major lawsuits against Suno, the company maintains that its models were trained on publicly available music files and metadata under the principle of fair use.