One week after Suno filed to keep the exact size of its AI training dataset sealed in the US District Court for the District of Massachusetts, UMG and Sony Music Entertainment pushed back. In a memorandum of law filed Friday (June 5), the two remaining plaintiffs in the nearly two-year-old copyright case urged Judge F. Dennis Saylor IV to reject Suno’s impoundment request, arguing the AI company cannot overcome the “strong presumption of public access” that attaches to court pleadings. The number at issue, which the labels’ filing calls the “Model Training Figure,” represents the total count of audio files the labels allege Suno used to build its generative AI model. UMG and Sony initially redacted it from their own court papers in an “abundance of caution” while reserving the right to contest sealing. Friday’s filing makes clear they intend to do exactly that.
The labels’ core argument is that Suno has already undermined its own confidentiality claim. In its answer to the complaint, Suno admitted that building its service “required showing the program tens of millions of instances of different kinds of recordings.” Having made that public acknowledgment, UMG and Sony argued, Suno “cannot credibly claim that disclosing a slightly more precise count of the audio files that its AI model trained on would reveal anything competitively sensitive.” The competitive harm theory offered by Suno CTO Georg Kucsko in his declaration, that rivals could use the figure to benchmark their systems, infer aspects of Suno’s training approach, and optimize their own models, was dismissed by the labels as “speculative” and legally insufficient. “Stripped of its conclusory assertions of competitive harm, there is no cognizable reason to shield from public view how many sound recordings, including Plaintiffs’ recordings, Suno copied to build its service,” the filing states.
The labels also moved to unseal a second figure: the count of recordings they identified in Suno’s training data as ones they had specifically provided to Audible Magic for fingerprinting. UMG and Sony argued Suno had effectively waived any objection to disclosing that number by failing to make any argument for sealing it in their original impoundment motion. “Contentions that are not developed or supported in a party’s submissions are deemed waived,” the filing states, citing prior rulings from the District of Massachusetts. The labels characterized both figures as directly relevant to Suno’s fair use defense because they “speak directly to the nature and extent of Suno’s copying,” one of the four statutory factors courts weigh in fair use analysis. How much a defendant copied, and from whom, is not peripheral to a copyright case: it is central to it.
The sealing dispute is running alongside a second front that has opened in the case simultaneously. On May 21, UMG and Sony filed a motion to amend their complaint to add 61,026 recordings identified through Audible Magic’s audio fingerprinting process after Suno refused to voluntarily disclose what recordings it had ingested. The original complaint, filed in June 2024 by the RIAA on behalf of all three majors, asserted 560 works. Suno opposed the amendment motion on June 4, arguing the labels had “unduly delayed” and that expanding the case at this stage would deny Suno a timely ruling on its fair use defense. A key summary judgment hearing on Suno’s fair use defense is currently scheduled for July 2026, and Suno’s argument is that adding more than 61,000 additional works this close to that hearing would reset the litigation clock in a way that prejudices its position. Brunomarsallegiantstadium
The parallel dispute over Udio’s training data size in the Southern District of New York, where Sony Music is the sole remaining plaintiff, is playing out under nearly identical conditions, with Udio’s legal team seeking the same confidentiality protections Suno is pursuing in Massachusetts. The convergence of the two cases on the same question, how much copyrighted material was ingested and whether the public has a right to know, reflects how central that number has become to the industry’s broader effort to establish what fair compensation for AI training actually looks like. As Suno closed a $400 million funding round at a $5.4 billion valuation this week and the AFM simultaneously filed its own lawsuit against UMG and Warner for cutting session musicians out of AI licensing proceeds, the gap between the financial stakes and the legal transparency available to the public has never been more visible. Fact discovery in the Suno case is scheduled to close June 26, though both parties have discussed pushing the deadline into August.
