Leaks Reveal Suno Fed Thousands of Hours of Deezer, YouTube and Pond5 Data Into Its AI
A hacker reportedly breached Suno, an AI music platform, and leaked source code plus internal logs that show how its training data was assembled. The files, reviewed by 404 Media, list large-scale scraping from YouTube Music, tagged YouTube tracks, Pond5, Deezer, and Genius-related material, and mention plans to download about 1 million hours of podcast audio via RSS. One log tracked more than 2 million YouTube Music clips. The leak strengthens music-industry claims that Suno trained on copyrighted songs pulled from YouTube. Suno says the incident was limited, involved outdated source code, and did not require customer notices; the hacker claimed access to some customer data, which Suno denies was compromised. Suno had already disclosed in broad terms that its training set included tens of millions of publicly available music files. The leak adds detailed evidence while Suno’s copyright lawsuit with Sony and UMG continues.
