Hack reveals Suno scraped millions of songs for AI training

A hack of the artificial intelligence music generation company Suno has revealed detailed source code allegedly showing the company systematically scraped millions of copyrighted songs and lyrics from major streaming platforms and music libraries to train its AI models. The breach, which the company says occurred in November 2025, exposed internal files that quantify the vast scale of the data collection, fueling ongoing legal battles and artist criticism over the ethics of AI training.

According to data from the hack provided to media outlets, the source code includes instructions and logs indicating Suno scraped at least 2,013,545 music clips from YouTube Music, amounting to over 113,000 hours of material. The files also list extensive scraping from other services, including 62,117 hours from the stock music library Pond5, 12,287 hours from the streaming service Deezer, and 17,615 hours from the lyrics site Genius. Additional sources reportedly included libraries such as Jamendo, Freesound, and the International Music Score Library Project. The hacker, using the pseudonym Ellie.191, stated they accessed the data by compromising an employee's credentials.