Sony Music Entertainment has urged a federal court to reject Udio‘s attempt to keep secret the total number of audio files it used to train its generative AI music model.
In a filing on Monday (June 8) in the US District Court for the Southern District of New York, which you can read here, the labels argued that the figure should be made public.
The number, which Udio‘s lawyers call the “Training Data Number,” is the single piece of information the AI company is fighting to redact from Sony‘s recent filings.
Sony’s filing mirrors a brief filed three days earlier in the parallel case against Suno, where Universal Music Group and Sony are urging a Massachusetts federal court to reject Suno’s bid to seal the size of its own training data.
Sony Music said the figure belongs on the public record because it appears in the labels’ proposed amended complaint, where the presumption of public access is at its strongest.
The plaintiffs also argued that the number “is necessary to understand the full extent of Udio‘s copying, which bears on Udio‘s fair use defense.”
“Udio‘s argument rings especially hollow because the Training Data Number is a lone aggregate number that says nothing about how Udio built its models or what it did with any sound recording.”
Sony Music’s filing
“Udio‘s argument rings especially hollow because the Training Data Number is a lone aggregate number that says nothing about how Udio built its models or what it did with any sound recording,” the filing states.
“It simply provides context for the scale of Udio‘s conduct, which the public is entitled to see,” the plaintiffs added.
Udio filed its motion to seal on June 1, telling the court that the figure is commercially sensitive.
The company argued that if competitors learned the size of its training dataset, they could “develop and refine competing products more rapidly and at a lower cost,” and could “extrapolate commercially sensitive insights about Udio‘s training methodology and strategy, thereby obtaining an unfair competitive advantage.”
In a declaration, Udio Co-Founder and CEO Andrew Sanchez said the size of a training dataset “is closely guarded because it reveals strategic business choices regarding model architecture and product development.”
Sony told the court those concerns are “speculative” and “conclusory,” and fall short of the standard required to seal a judicial document.
“Stripped of its conclusory assertions of competitive harm, there is no cognizable reason to shield from public view how many sound recordings Udio copied to build its service.”
Sony Music’s filing
“Stripped of its conclusory assertions of competitive harm, there is no cognizable reason to shield from public view how many sound recordings Udio copied to build its service,” the plaintiffs wrote.
The filing said Udio‘s motion “seeks to conceal the full extent of its copying to create its training database.”
The dispute is the latest turn in a sealing fight running alongside Sony‘s effort to expand the case.
On June 3, Judge Alvin K. Hellerstein vacated an earlier order that had granted Udio‘s request to seal, inviting both sides to brief the question in the ordinary course.
The Training Data Number is distinct from the 30,442 copyrighted works that Sony is seeking to add to its complaint.
The labels say they identified those recordings after gaining access to Udio‘s training data in discovery, under a protocol that designated the material “HIGHLY CONFIDENTIAL – ATTORNEYS’ EYES ONLY.”
Udio has said it is not seeking to seal those specific recordings, telling the court that “the sole information Udio seeks to impound is the total number of audio files allegedly used to train its model.”
The request mirrors a move by rival AI platform Suno, which asked a federal court in Massachusetts to seal the size of its own training data in the labels’ separate case against it.
All three major record companies – Sony, Universal Music Group and Warner Music Group – sued Udio and Suno in June 2024, accusing both of “mass infringement” of copyright.
UMG settled with Udio in October 2025 and Warner followed in November 2025, leaving Sony as the sole remaining major-label plaintiff in the case.
In its answer to the amended complaint in April, Udio admitted that it “obtained audio data from YouTube for use as training data,” while maintaining that its use of copyrighted music amounts to “fair use.”
Both sides are heading toward motions for summary judgment focused on the fair use question, with document production scheduled to close on June 26, 2026.
Sony is represented by law firm Hueston Hennigan, while Udio is represented by Latham & Watkins.Music Business Worldwide
