Hack Reveals Suno Used YouTube and Deezer Data to Train AI Music Models

Variety reports that a hacker accessed internal source code from Suno AI in November 2025, revealing instructions that directed the scraping of large volumes of data from YouTube Music, Deezer, Genius, and several stock music libraries. The breach also exposed partial customer information, including emails, phone numbers, and Stripe payment details.

The leaked code reportedly showed that millions of song clips and metadata from these sources were used to train Suno's AI music generator. Specific datasets included over two million clips from YouTube Music and more than one hundred thousand hours of associated audio and lyric sources.

A Suno spokesperson said the incident was contained soon after discovery and affected only outdated code no longer in use. The company stated that its models are trained on publicly accessible music files and related data from across the internet. It confirmed that no full credit card numbers were exposed and said the breach did not require customer notification.

Suno has previously acknowledged using publicly available music to train its systems, asserting that the practice falls under fair use protections. The company remains involved in copyright disputes with major record labels over data used in training its models.