Thanks to a breach in November 2025 it appeared how huge the music material Suno collected for her AI models, with over two million clips only from YouTube Music.
A hacker allegedly violated the AI Suno music production platform in November 2025 gained access to the data file behind the company's models, throwing light on the hardware scale Suno seems to have drawn from streaming platforms. According to a 404 Media report, the archive contains a number of artists' works, offering a rare look inside a system that usually functions as «black box».
The hacker allegedly acquired a Suno source code covering the period 2023 to 2024, including scraping instructions and data on the extent of the material drawn by the model. The code suggests that Suno collected material from the Genius lyrics platform, YouTube Music, streaming Deezer platform, as well as from music stock libraries such as Freesound and International Music Score Library Project.
The sizes are impressive. A file about YouTube Music indicated draw 2,013.545 music clips from the platform. Another set of data measured the material in hours instead of records: 113,879 hours of music from YouTube Music, 17,615 from Genius and 62,117 hours from the Royalty-free music website Pond5, among many more sources.
A Suno spokesman confirmed the violation at Gizmondo. «In November 2025 we found that Suno had suffered a limited safety incident, which was quickly restricted», stated the representative. «At that time, we immediately conducted an investigation and verified that the incident was primarily a source code that is no longer used in Suno.»
The breach allegedly also included email or customer phone numbers, as well as payment details via Stripe. Suno argued that «did not leak sensitive personal data» and specify: «Importantly, Suno does not have access to customer credit card full numbers in Stripe.» The company did not inform the affected customers individually, considering that «individual notifications were not mandatory under applicable data protection legislation».
As to the training material itself, Suno argues that the breach did not reveal anything new. «Our goal has always been to help people create original new music rather than copy the work of others. This is why we develop our models around what we call "Original Creation, By Design"», said the representative, adding that the company does not use artists' names as a category of education metadata and that it has incorporated filters that prevent users from using artists' names, song titles or albums as prompts.
Suno faces many lawsuits for her model. In a lawsuit filed against her by large records, including Universal Music Group, Capitol Records, Atlantic Records, Warner Music and Sony Music, the plaintiffs argue that the prompt «1954 rock and roll billy Haley comets» It produced a result that allegedly directly copied Bill Haley's style and melody, which contradicts the firm's claims on safeguards.
The company considers that the use of the material was legitimate. «Suno's AI models have been trained in music records that are freely available to the public and related metadata, which are accessible to third party websites on the open internet», the representative said. In court documents Suno has admitted that her educational data «include virtually all decent quality music files accessible on the open internet»But citing the frame of the fair use.
Whether this legal defence remains an open question, as so far it has been issued minimum and conflicting decisions on the matterSuno chose not to expect clear jurisprudence and developed its model by calculating that it will either be justified on the basis of fair use or will pay any fine imposed on it. Thanks to the violation, what is now becoming clearer is the scale of the material that was drawn.
-
1
-
2
