On July 20, 2026, Sony Music Entertainment and a roster of its labels — Arista, LaFace, Zomba, Ultra, Volcano, and Alamo among them — filed a second federal copyright lawsuit against Uncharted Labs, the company behind the AI music generator Udio. The complaint alleges that Udio copied Sony's sound recordings — without permission, license, or payment — to train the models that let anyone type a prompt and get back a finished song. The case is Sony Music Entertainment v. Uncharted Labs, Inc., No. 1:26-cv-06120, filed in the Southern District of New York under the Copyright Act (17 U.S.C. § 101 et seq.). Within roughly a day it drew coverage from more than 30 outlets — The Verge, Billboard, Variety, the Los Angeles Times, and the music-industry trades among them — making it one of the fastest-spreading new filings on the IP Feed.
What sets this one apart is the word again. This is not Sony's opening move against Udio but its second suit, and the sequel is bigger than the original. According to the coverage, Sony brought a fresh complaint over more than 30,000 additional recordings that it was not permitted to fold into its earlier case — the 2024 action that the major labels, coordinated through the recording industry, filed against both Udio and rival generator Suno. The new docket was accepted as related to that original suit, No. 1:24-cv-04777, and both now sit before Judge Alvin K. Hellerstein in Manhattan. Some reports pegged the theoretical statutory exposure across those 30,000-plus recordings in the billions of dollars — a ceiling, not a forecast, but a measure of the stakes.
What the lawsuit says
The core allegation is by now familiar from the wider wave of AI-training disputes, but the medium here is music rather than text or images. Sony says Udio ingested tens of thousands of its copyrighted sound recordings and used them, without authorization, as training data — the raw material a generative model learns from in order to produce new audio on demand. Recorded music, the labels argue, is a valuable proprietary asset built on decades of investment in artists, studios, and catalog, and Udio built a commercial product on top of it while paying nothing for the input.
Suits in this wave typically go beyond the raw copying claim. Rights holders have argued that generative systems can reproduce protected material in close imitation — melodies, vocal timbres, and production styles recognizable as the originals — and that the resulting flood of machine-made tracks competes directly with the human catalog for streams, sync placements, and listener attention. The scale is part of the argument: a model trained on 30,000 recordings, the labels contend, is not sampling but wholesale copying dressed up as learning.
The procedural backstory matters too. The reason this is a separate lawsuit at all is that the court in the first case declined to let Sony keep expanding it, capping the recordings at issue there. Rather than lose the rest of its claims, Sony filed them anew. The defendant has not yet answered the new complaint, and no court has ruled on its merits. What follows is the legal question the filing raises — not a prediction of how it comes out.
The hard part: fair use
Every one of these cases runs into the same doctrine, and it is the one that will decide them: fair use. The defense's central argument is that training a model on copyrighted works is transformative — that the system does not store and replay the recordings but learns statistical patterns from them to generate something new. AI companies point to a line of cases where large-scale copying was held to be fair use because the output served a fundamentally different purpose from the original.
The labels attack that framing on the fourth fair-use factor — the effect on the market for the original work. Their argument is that AI music is not a neutral tool but a substitute good: a system trained on their catalog that then produces songs competing for the same playlists, licensing budgets, and ears. If a court accepts that the output displaces the originals in the market, the transformative-use defense weakens considerably. Music sharpens the point in a way text does not — a generated track can occupy the exact commercial slot a licensed recording would have filled.
Courts weigh four factors — the purpose and character of the use, the nature of the work, the amount taken, and the market effect — and no single factor is dispositive. The unsettled question, the one no appellate court has yet resolved cleanly for generative AI, is how those factors apply when the amount taken is effectively an entire catalog and the "new" use competes for the same audience. That is the gap Sony is trying to widen, and the one Udio will work to close.
Why it matters beyond one lawsuit
For anyone who owns rights in creative work, this litigation is about far more than one AI startup. It is one of the pressure points that will determine whether training on copyrighted material requires a license at all — and, if it does, what a market for those licenses looks like. Some rights holders have already struck licensing deals with AI companies; others, like Sony here, are litigating instead. The outcomes will shape which path becomes the default for music, and by extension for film, television, and every other catalog business built on recorded IP.
There is a quieter lesson for rights holders too. Cases like this turn on the ability to prove exactly what was created, who owns it, and when — the same provable chain of authorship and ownership that underpins the value of any IP asset. When the fight is whether your work was taken and whether its market was harmed, a clean record of what you own and how it has been exploited is not paperwork; it is leverage. For independent creators without a major label's litigation budget, that record is often the only leverage there is.
We'll track this docket as it develops. The first real signals will come as the defendant files its response and the shape of its fair-use defense comes into view, as the new case is coordinated with the original before Judge Hellerstein, and as the court begins to test whether "training" on 30,000 recordings looks more like transformation or like copying. Follow the filings, counsel, and coverage on the case page.
This post is editorial commentary on public court filings and news coverage, not legal advice. The allegations described are unproven, and the defendant has not yet responded on the merits. Damages figures reflect statutory maximums cited in press coverage, not any court finding.