Suno Hack Shows How YouTube Music, Deezer and Genius Data Trained AI Music Generator’s Models
Suno has long acknowledged that its AI music generator relied on the scraping of millions of songs available across the internet, but a new hack reveals...
Key points
- Suno has long acknowledged that its AI music generator relied on the scraping of millions of songs available across the internet, but a new hack reveals just how the company pulled from streaming services and websites such as YouTube Music, Deezer and Genius to power its product — all while user information remained vulnerable.
- A report in 404 Media on Wednesday, relying on data a hacker provided to the outlet, showed the instructions in the company’s source code had it scrape files from “genius_hq, youtube_music, freesound, jamendo, imp, deezer,” with the stock music libraries Freesound, Jamendo and the International Music Score Library Project among the other sources scraped.
- “As we have stated in public filings and disclosures, Suno’s AI models have been trained on publicly available music files and related metadata accessible on third-party websites on the open Internet,” the spokesperson said.
- One file focused on YouTube Music noted it took in “2,013,545 music clips” at the time it was last updated, while another file’s comments about Suno’s datasets included “113,879 hours of youtube_music,” “17,615 hours of genius_hq,” “410 hours of free sound,” “19,514 hours of imslp,” “3,726 hours of jamendo,” “62,117 hours of pond5_music,” “12,287 hours of deezer,” “152,162 hours of ytm_tagged” and “103 hours of musescore_lyrics,” according to the 404 Media report.
What happened
Suno has long acknowledged that its AI music generator relied on the scraping of millions of songs available across the internet, but a new hack reveals just how the company pulled from streaming services and websites such as YouTube Music, Deezer and Genius to power its product — all while user information remained vulnerable. A report in 404 Media on Wednesday, relying on data a hacker provided to the outlet, showed the instructions in the company’s source code had it scrape files from “genius_hq, youtube_music, freesound, jamendo, imp, deezer,” with the stock music libraries Freesound, Jamendo and the International Music Score Library Project among the other sources scraped. “As we have stated in public filings and disclosures, Suno’s AI models have been trained on publicly available music files and related metadata accessible on third-party websites on the open Internet,” the spokesperson said. One file focused on YouTube Music noted it took in “2,013,545 music clips” at the time it was last updated, while another file’s comments about Suno’s datasets included “113,879 hours of youtube_music,” “17,615 hours of genius_hq,” “410 hours of free sound,” “19,514 hours of imslp,”...
Related on this blog
- Paramount Blasts States’ Lawsuit Aimed at Blocking Warner Bros. Deal, Claims Litigation Will ‘Shield’ Netflix and Tech Companies From ‘Much-Needed Competition’
- Hack suggests AI music generator Suno scraped YouTube for training data
- Death Threats, Doxxing and AI: How Animator Jorge Gutiérrez Killed His Own Show in 36 Hours (EXCLUSIVE)