Suno breach claims put training data sourcing under scrutiny
Between July 15 and 17, multiple accounts circulated claims that AI music company Suno had been breached, with part of its source code and materials related to training data leaking out. The story drew attention not only as a security incident, but because the leaked material was framed as evidence about where Suno’s training corpus came from and whether it was properly licensed.
Key details
Based on leaked code and circulating materials, xiaohu said Suno’s training-data-related code indicated scraping of roughly 380,000 hours of audio from platforms including YouTube Music, Pond5, and Deezer. The figures highlighted in posts were 113,879 hours from YouTube Music, 62,117 hours from Pond5, and 12,287 hours from Deezer. Polymarket and TuhinChakr also relayed claims that the materials extended beyond music to lyrics and podcasts, involving sites such as Genius.
Alleged access path and possible impact
Posts relayed by evilsocket and nptacek pointed to the Shai-Hulud worm as a possible route of compromise. In a separate post, xiaohu said the attacker may have entered Suno’s internal systems through an employee account and allegedly accessed user email addresses, phone numbers, and Stripe payment information. The posts did not provide a clearer scope of affected users or systems.
Controversy and uncertainty
The broader debate focused on whether the training data had been licensed, where platform scraping boundaries should lie, and what the leaked materials might imply for compliance in generative music. Gary Marcus, in a repost, noted that the leaked inventory was detailed down to hour counts and also mentioned that Suno had already been sued by major record labels. Still, the claims in this cluster mostly came from leaked code and secondary retellings, and no Suno official response appeared in the provided posts, leaving the completeness and authenticity of the leaked materials unverified.
2026-07-15 ~ 2026-07-17 · 7 related posts
- Suno Hacked: Training Data and Source Code Allegedly Leaked — TuhinChakr · 2026-07-15
- [source] Suno Accused of Scraping Massive Music Datasets — Polymarket · 2026-07-16
- Suno Breached: Training Data Leaked — evilsocket · 2026-07-16
4 near-duplicate retellings: nptacek · GaryMarcus · xiaohu · xiaohu