Breaking

Hack of AI Music Startup Suno Exposes Source Code Detailing Scraped Training Data From YouTube, Deezer and Genius

July 16, 2026 at 10:23 EDT

Security
Media Generation
Policy & Regulation

A breach of the AI music company Suno has surfaced internal source code that appears to catalogue in unprecedented detail how the startup assembled its training set, drawing on tens of thousands of hours of audio scraped from YouTube Music, Deezer, Genius and other services. The intrusion, reported this week, has reignited the fight over how generative-music systems acquire their data and whether such practices fall within U.S. copyright law.

November 2025 Breach · Suno AI

Leaked Source Code Exposes How an AI Music Startup Scraped Its Training Set

A self-propagating supply-chain worm stole an employee's credentials, surfacing internal code that catalogues in unprecedented detail the sources, methods and scale behind Suno's training audio — drawn from YouTube Music, Deezer, Genius and more.

113,879

hours from YouTube Music — over 2 million clips

~1M

hours of podcasts planned via PodcastIndex

distinct services named as scraping sources

Audio hours per source, as accounted for in the leaked code

Column height is proportional to hours ingested

113,879

62,117

17,615

12,287

3,726

YouTube
Music

Pond5

Genius

Deezer

Jamendo

The intrusion

Shai-Hulud worm → Stolen credentials → 2023–24 code + records → Given to journalists

Exposed data reportedly included emails, phone numbers and some Stripe payment information.

Suno's position

Breach contained quickly, the exposed code was largely obsolete, and no sensitive personal data was compromised — so individual notifications were deemed unnecessary. Training material described as publicly available files from the open internet, argued as fair use.

Why it matters

A rare primary record of the exact sources, methods and scale behind training claims — arriving amid active RIAA litigation. For creators, the source list raises a pointed question: was their work ingested?

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.