Crypto
Home›Crypto›Market Structure›Suno source-code leak details large AI training data p…
Suno source-code leak details large AI training data pull from music sites
The leaked materials say the training set included more than 113,000 hours from YouTube Music, plus stock audio from Pond5 and Deezer, and it also describes plans to download about 1 million hours of podcast audio via RSS feeds.
Decrypt reports that leaked source code from an alleged 2025 breach at AI music platform Suno documents how its training library was assembled, including scraping instructions and internal logs tied to ingestion pipelines.
According to the leak as reviewed by Decrypt and first reported by 404 Media, the dataset included 113,879 hours from YouTube Music, 62,117 hours from Pond5, and 12,287 hours from Deezer, along with additional material from tagged YouTube tracks and a dataset labeled genius_hq linked to Genius.
The files also describe plans to download roughly 1 million hours of podcast audio via RSS feeds, and one internal log reportedly recorded about 2,013,545 music clips tied to YouTube Music ingestion.
The hacker also claimed access to records for hundreds of thousands of customers, including emails, phone numbers, and Stripe-related data, while Suno disputed that sensitive personal information was compromised and said it identified the incident in November 2025 and considered it limited.