Musicbusinessworldwide.com
AI music startup Suno is pushing back against efforts to publicly reveal the scale of the dataset used to train its music-generation technology, arguing that disclosure could damage the company competitively as its legal battle with major record labels intensifies.
In a recent filing submitted to federal court in Massachusetts, Suno requested that one specific detail remain sealed from public view: the precise number of audio files allegedly used to train its generative AI model. The dispute is unfolding within the broader copyright infringement lawsuit brought by Universal Music Group and Sony Music Entertainment, both of which accuse the company of using copyrighted recordings without authorization.
Suno’s legal team emphasized that the company is not attempting to hide which recordings the labels claim were included in the training process, nor is it seeking to block references to the tens of thousands of additional songs the plaintiffs want added to the case. Instead, the company says its request is narrowly focused on protecting a single numerical figure that it considers highly sensitive business information.
The company’s Chief Technology Officer, Georg Kucsko, argued in a supporting declaration that the size of an AI training entity can reveal strategic insights into how a model is developed and optimized. According to Suno, competitors could potentially use that information to benchmark their own systems, infer technical approaches and accelerate rival products using knowledge derived from Suno’s internal operations.
The filing comes after Inner City Press challenged the sealing request, arguing that information surrounding the company’s training practices is central to public understanding of the copyright claims. Suno strongly disputed that characterization, stating that the broader legal questions in the case can still be understood without exposing confidential data tied to its development strategy.
The lawsuit itself continues to expand. Universal and Sony recently sought permission to add more than 61,000 additional recordings to their complaint after identifying alleged matches through audio fingerprinting technology. The labels claim Suno trained its systems on millions of copyrighted tracks while refusing to fully disclose the material used.
About the Author
