Suno caught scraping millions of songs from YouTube, Genius, and Deezer

Hacked files obtained by 404 Media expose that AI music startup Suno scraped millions of songs from platforms like YouTube Music, Deezer, and Genius to train its generative models. Source code from 2023-2024 includes scraping instructions targeting over 2 million YouTube Music clips, plus extensive datasets from Deezer, Jamendo, Freesound, and the International Music Score Library Project. The data also reveals Suno used a third-party service called Bright Data to scrape YouTube and specifically searched for a cappella versions for vocal-only training. Additionally, code shows plans to download roughly one million hours of podcasts via PodcastIndex.

These revelations come amid active litigation: the RIAA sued Suno for using copyrighted music without permission, and an amendment alleged Suno deliberately circumvented YouTube's protections (stream-ripping). Suno has publicly argued that training on publicly available internet files falls under fair use. The same hack also exposed customer data—emails, phone numbers, and Stripe payment details—though Suno claims it determined individual notifications were not required under privacy laws. The incident highlights the ongoing tension between AI training practices and copyright enforcement in the music industry.