Suno AI Music Built on Scraped YouTube Songs
Suno AI music was trained on 2M+ scraped YouTube songs.
A Hack Exposes the Truth
A hack of Suno AI music exposed its source code, revealing it scraped millions of songs from YouTube, Deezer, Genius, and Pond5 to train its song generator.
July 15, 2026 – 9:30 pm
Image by: Suno AI
For years, musicians have accused AI song generators of using their work without permission. A hacker has now lifted the lid on how this happens.
Leaked source code from Suno, one of the largest AI music tools, tells the story. It trained its model by scraping millions of songs and lyrics from across the internet.
404 Media obtained the data from a hacker who breached the company.
The haul is substantial. One file, labelled "youtube_music," contained over 2 million clips. Others list tens of thousands of hours pulled from Deezer, Genius, and Pond5. In total, it adds up to decades of recorded music.
Ripping Vocals and Bypassing Defences
The code was specific about its targets. To capture clean voices, it sought a cappella versions of songs on YouTube. To bypass YouTube’s defences, Suno used a proxy firm called Bright Data. It also trawled 420,000 podcasts, seeking around one million hours of speech.
The EU Tech Connection
None of this is new. The record labels have long made similar allegations. In court, Suno admitted training on "essentially all music files of reasonable quality" on the open web. However, the leak provides a detailed look at the machinery behind that statement.
The timing is significant. Some labels have struck licensing deals with AI firms rather than litigating. Sony is still in court, awaiting a pivotal fair-use ruling this summer. Meanwhile, artists continue to express dissatisfaction with these deals, arguing they do little to benefit them.