Suno has asked a federal court to keep secret the exact number of audio files it used to train its AI music generation model – arguing that disclosing the figure would give rivals a competitive advantage.
The AI music company filed its statement in the US District Court for the District of Massachusetts on Friday (May 29), responding to a press opposition from Inner City Press (read here) that had urged the court to unseal training data evidence in the copyright infringement lawsuit brought by Universal Music Group and Sony Music Entertainment.
Suno said in the court filing, which you can read here, that it is seeking to impound just one piece of information across the labels’ recently filed court papers: a single figure representing the total volume of audio files the plaintiffs allege Suno used to train its generative AI model.
Suno also argued that Inner City Press‘s request is now moot – because the court ordered Suno to make a specific good-cause showing for impoundment, and Suno says it has done so.
“Because the Court has ordered Suno to make such a showing, and Suno has done so, Inner City Press’s request is moot,” the filing stated.
The filing makes clear that Suno is not attempting to seal the identity of any specific recordings the labels claim were in its training data, nor the 61,026 additional copyrighted works UMG and Sony are seeking to add to their lawsuit.
“Suno does not seek to impound any particular recording Plaintiffs allege was included in the training data,” the company’s lawyers wrote.
“Rather, the sole information Suno seeks to impound is the total number of audio files allegedly used to train its model.”
The filing was accompanied by a declaration from Georg Kucsko, Suno‘s Chief Technology Officer, who stated that the figure “is not publicly available and has not been disclosed outside of this litigation.”
“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, including with respect to the exact volume of data required to build a high-quality generative AI model.”
Georg Kucsko, Suno
Kucsko argued that public disclosure “would risk significant competitive harm to Suno.”
“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, including with respect to the exact volume of data required to build a high-quality generative AI model,” Kucsko stated in the declaration, which you can read in full here.
“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 labels’ court papers already refer publicly to “millions” of sound recordings in Suno‘s training data.
Suno acknowledged that it is not seeking to impound those more general references, only “a handful of references to the exact number of files in its training data set.”
The filing also pushed back directly on arguments made by Inner City Press reporter Matthew Russell Lee, who filed a letter on May 22 opposing the impoundment.
“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.”
Georg Kucsko, Suno
Lee had argued that the sealed materials “address what music Suno used to train its AI system” and that the information “goes to the heart of the copyright infringement claims.”
Suno called that 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.
“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.”
Suno’s statement in support of motion for impoundment
Suno‘s lawyers argued that the plaintiffs’ infringement claims are based on specific works they have identified – not on the total size of the training data set – and that those claims and Suno‘s defenses “are well-documented in the parties’ public filings.”
The filing noted that the court has previously granted motions to impound information about Suno‘s training data “due to the same competitive risks of disclosure currently at issue.”
It also cited sealing decisions in other AI copyright cases, including The New York Times v. Microsoft, Kadrey v. Meta Platforms, and Concord Music Group v. Anthropic.
The plaintiffs do not oppose Suno‘s impoundment request, but have “reserved the right to contest the propriety of sealing this material at a later date,” according to the filing.
Suno also signaled that it intends to oppose the labels’ underlying motion to amend the complaint – which seeks to add the 61,026 additional recordings.
UMG and Sony filed the motion to amend on May 21 after using Audible Magic audio fingerprinting technology to identify their recordings within Suno‘s training data.
The labels say discovery revealed that Suno used “millions” of their copyrighted sound recordings to train its AI models, and that Suno “continued to refuse to identify the sound recordings it had used to train.”
The original complaint, filed in June 2024, asserted 560 copyrighted works.
Warner Music Group (formerly a co-plaintiff) agreed a settlement with Suno in November 2025 and struck a licensing deal with the company.
UMG and Sony remain active plaintiffs, and their licensing negotiations with Suno have reportedly stalled.
The two remaining labels have also been fighting to obtain the terms of Suno’s settlement with Warner Music – a request that a federal magistrate blocked in an April 6 discovery ruling.
The most recent scheduling order sets a deadline for dispositive motions of January 8, 2027.Music Business Worldwide
