Suno Is Trying to Keep One Number Secret. That Number Could Define the AI Copyright Case.

In a filing submitted Friday (May 29) to the US District Court for the District of Massachusetts, Suno asked the court to keep sealed a single data point from the copyright infringement lawsuit brought against it by UMG and Sony Music Entertainment: the exact total number of audio files the labels allege Suno used to train its generative AI model. The company is not asking to seal the identity of specific recordings, nor the 61,026 additional copyrighted works UMG and Sony are seeking to add to the case through a motion to amend filed May 21. It is asking to protect one figure, which it describes as competitively sensitive business intelligence that has never been publicly disclosed and whose release could meaningfully benefit rival AI music companies.

The filing was triggered by a May 22 letter from Inner City Press reporter Matthew Russell Lee, who urged the court to unseal training data evidence, arguing the information “goes to the heart of the copyright infringement claims” and that the public has a right of access to materials in a case of this significance. Suno’s lawyers pushed back directly, calling Lee’s characterization false. “Inner City Press’s sweeping and generalized statements about the right of public access and the importance of this litigation are simply divorced from the reality of what Suno seeks to impound: a small number of references to a single, competitively sensitive figure,” the filing stated. Suno also argued the request is now moot because the court had ordered Suno to make a specific good-cause showing for impoundment, and the company says it has satisfied that standard.

The competitive harm argument is laid out in detail in a declaration from Suno Chief Technology Officer Georg Kucsko, who stated that the figure “is not publicly available and has not been disclosed outside of this litigation.” Kucsko’s reasoning is direct: “In the rapidly evolving and highly competitive generative AI market, the size of a company’s training corpus reflects technical development decisions and strategic judgments concerning model design and performance. If competitors obtain access to the Model Training Figure, this could allow them to benchmark their own systems against Suno’s model, infer aspects of Suno’s training and development approach, and potentially optimize their models to unfairly compete with Suno’s by leveraging Suno’s confidential business information.” The court has previously granted motions to impound other information about Suno’s training data on the same competitive risk grounds, and Suno’s lawyers also cited sealing decisions in the Concord Music Group v. Anthropic case, The New York Times v. Microsoft, and Kadrey v. Meta Platforms as supporting precedent.