Hack Reveals Suno AI Music Generator Scraped YouTube, Deezer, and Genius
404 Media Jason Koebler ● Covered by 2 sources
A hacker breached Suno and found it scraped millions of songs from YouTube Music, Deezer, Genius, and more to train its AI. The leak also exposed data on hundreds of thousands of paying Suno users, including Stripe payment info.
Based on reporting by 404 Media, Jason Koebler — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
For months, Suno has been fighting record labels in court over claims it trained its AI on copyrighted music without permission. Now someone has actually shown the receipts.
A hacker who broke into Suno's systems handed 404 Media a trove of internal source code and scraping logs dating from 2023 and 2024. The files lay out, in granular detail, exactly where the company pulled its training data from: YouTube Music, Deezer, Genius, Pond5, Jamendo, Freesound, the International Music Score Library Project, and podcast RSS feeds. One file logs 2,013,545 music clips ingested from YouTube Music alone. Another tallies up the hours by source — 113,879 hours from YouTube Music, 152,162 hours from a tagged YouTube Music set, 62,117 hours from Pond5, 19,514 from IMSLP, 17,615 from Genius lyrics, 12,287 from Deezer, and smaller chunks from Freesound, Jamendo, and MuseScore. Add it up and you're looking at what's probably decades of recorded audio.
None of this is exactly shocking on its own. Suno already admitted in court filings that it trained on "essentially all music files of reasonable quality" it could find on the open internet, tens of millions of recordings by its own count. The company's legal defense hinges on fair use, and one of the RIAA's lawsuits against it has already settled. But admitting something in the abstract during litigation is different from having the actual file names and clip counts leak out. The RIAA specifically accused Suno of ripping tracks straight off YouTube, and this leak backs that claim up with numbers instead of allegations.
The bigger problem for Suno right now might not even be the training data. The same hacker also got into user records and Stripe payment details for hundreds of thousands of customers, which is a separate and more immediate mess for a company already under legal fire. A breach that confirms your training practices while also exposing your paying users' financial data is the kind of week most startups don't recover from cleanly.
What this leak really does is give critics and plaintiffs something concrete to point to. Court admissions can be argued over, framed, softened by lawyers. Scraping logs with exact clip counts from named platforms are much harder to spin.
My take — AI-written commentary, not fact-checked reporting
I don't think Suno is unique here, I think it's just the one that got caught with its logs showing. Every generative model trained on 'essentially all music files of reasonable quality' on the internet knew exactly what it was doing, fair-use argument or not, and this leak is going to make that defense a lot harder to sell to a jury.
Read more about this at: 404 Media