A data breach has revealed that Suno trained its artificial intelligence model by scraping millions of copyrighted songs and lyrics from YouTube Music, Deezer, and Genius. The hack exposed internal files showing the company accessed these platforms without permission, contradicting previous claims about how its training data was gathered. This incident provides concrete evidence of the material Suno actually used to generate its music outputs, an area the firm has kept opaque during public disputes.
The findings matter because they directly address ongoing legal challenges regarding copyright infringement. The Recording Industry Association of America and other rights holders have argued that Suno violated intellectual property laws by ingesting protected content. By confirming the scale of this data collection, the leak strengthens the case that the service operated outside fair use boundaries. It also highlights the difficulty of policing unlabelled training data in generative AI development.
- The exposed dataset included full audio tracks and associated lyrics.
- Suno has not publicly disclosed specific details about its training methodology.
- Multiple lawsuits are currently pending between the company and music industry groups.
