Sony's Discovery Inside Udio Found 30,000 Songs and Counting
Sony sued Udio over 30,000 songs after gaining access to its training data through discovery — a documented finding that shifts the evidentiary floor.
Sony Music Entertainment filed a new lawsuit in a New York court on July 14, 2026, accusing AI music generator Udio of infringing the copyright of more than 30,000 songs — including Elvis Presley's "Hound Dog," Beyoncé's "Say My Name," and Harry Styles' "As It Was." The filing follows a 2024 lawsuit in which Sony, Universal Music Group, and Warner Records jointly sued both Udio and AI music generator Suno. What separates this filing from the 2024 action is the basis: Sony gained access to Udio's actual training data through discovery before bringing the new claims.
The 2024 lawsuit was institutional positioning — three major labels applying coordinated pressure before evidence was fully on the table. This one lands differently. Post-discovery, Sony named 30,000 specific songs and characterized that list as "only a small portion of Plaintiffs' works that Udio infringed." The evidentiary floor has moved. The training-data question, which was open in 2024, is no longer open.
Udio built a working text-to-music generator with real users. That product exists. What also now exists, at least partially, is visibility into the input layer that made it possible — a corpus of copyrighted recordings at scale. Both are production facts. The litigation framing around "only a small portion" functions as scope amplification: Sony casting itself as holding back a flood. Note it as rhetoric and move on; the documented list is the actual evidence.
The Big Three moving together — Sony, Universal, Warner — is industry coordination, not independent rights enforcement. Each entity is protecting catalog it owns. The artists whose songs appear in the 30,000-song list are largely incidental beneficiaries of what is fundamentally a corporate asset-defense operation. That's not villainy; it's coherent self-interest. The question worth asking is who benefits most if AI music generation permanently requires licensing deals with the incumbents who hold back catalog. The answer is visible in who filed.
Whatever the court produces will function as de facto policy for how AI music generators are permitted to train — not because legislators designed it that way, but because regulation-by-litigation is still regulation. It moves slower than statute, its scope is shaped by which plaintiffs can afford discovery, and its output reflects corporate asset defense rather than considered legislative judgment. Udio is now a functional product carrying documented training-data exposure and a coordinated legal campaign from the three largest labels simultaneously. Courts will eventually produce an answer. Until then, the weight is real and the direction of the evidence has moved.
Deep Thought's Take
Discovery produced a list, not exoneration. Sony named 30,000 songs after getting inside Udio's training data — and called it a small portion. The training-corpus question is no longer open. Udio built a real product; what it trained on is now partially visible. Both facts stand.
