AI Music Startup Suno Admits to Training on Copyrighted Songs
Suno officially confirms it trains its AI model on copyrighted music but argues the practice is protected under the fair-use doctrine. The RIAA strongly disagrees, calling the action industrial scale infringement that threatens artists' livelihoods.
AI music startup Suno admits in a recent court filing that it trains its AI model using copyrighted songs, but the company claims this practice is completely legal under the fair-use doctrine. The filing comes as a direct response to a lawsuit filed by the Recording Industry Association of America (RIAA) in late June, which accuses Suno and fellow startup Udio of unauthorized use of protected music. While investors previously hinted that Suno lacked permission to use this material, this legal document represents the most explicit confirmation to date.
Suno CEO Mikey Shulman defends the company's methods by comparing AI training to a young musician learning to write rock songs by listening to the genre. In a blog post accompanying the court filing, Shulman emphasizes that the company simply trains its models on medium- and high-quality music found on the open internet, which naturally contains copyrighted materials owned by major record labels. He insists that the act of learning from existing art is not and has never been an act of infringement.
The RIAA firmly rejects this comparison, arguing that Suno only makes this concession because the lawsuit forces them to reveal facts they previously tried to hide. The industry group asserts that extracting the core value of an artist's life's work to create competing products is outright theft rather than fair use. Because this case challenges how traditional copyright doctrine applies to modern AI training, its final outcome sets a crucial precedent for the entire future of generative artificial intelligence.