Musicbusinessworldwide.com
Fresh information about AI music company Suno’s training methods could add new momentum to its ongoing copyright battle with major record labels after hacked internal source code reportedly revealed the platforms used to build its generative music models.
According to a report published by 404 Media, code obtained through a security breach references YouTube Music, Deezer, Genius, Jamendo and several other online services as sources for audio used to train Suno’s AI. The leaked files also allegedly detail the scale of the company’s data collection, including more than two million music clips pulled from YouTube Music and hundreds of thousands of hours of audio sourced from various platforms.
The disclosures come as Suno continues to defend itself against a copyright infringement lawsuit brought by Universal Music Group and Sony Music Entertainment, coordinated through the Recording Industry Association of America (RIAA). The labels argue that the company copied copyrighted recordings without permission to train its AI models, despite Suno arguing that using publicly available material for AI training qualifies as fair use under U.S. copyright law.
The leaked code reinforces one of the label’s central allegations, that Suno obtained recordings directly from YouTube. In a previous amended complaint, the RIAA accused the company of “stream ripping” music from the platform and bypassing technological protections designed to prevent unauthorized downloading.
According to 404 Media, the code also references tools used to locate a cappella recordings and proxy services that may have facilitated large-scale scraping. Additional datasets reportedly include audio from Pond5, Deezer, the International Music Score Library Project, Jamendo and podcast archives.
The legal stakes remain substantial. Earlier this year, Universal and Sony asked the court to expand the case from 560 allegedly infringed works to more than 61,000 recordings identified through audio fingerprinting. If approved, the move could dramatically increase the potential statutory damages sought by the labels.
Suno acknowledged in a statement that its AI models were trained using music files and metadata available across the public internet, consistent with disclosures it has already made in court.
About the Author
