Labels v. Udio

The allegations

The labels' complaint, filed by an RIAA-coordinated coalition on June 24, 2024, accused Uncharted Labs of copying copyrighted sound recordings at scale to train Udio, the generative music tool briefly famous for the viral "BBL Drizzy" track. The suit pled direct copyright infringement and sought statutory damages of up to $150,000 per work, plus an injunction. It landed the same week as a nearly identical complaint against rival startup Suno in Boston federal court, and the two cases have moved on parallel, closely watched tracks ever since.

The theory was straightforward: Udio's outputs echoed the phrasing, structure and style of specific, identifiable artists closely enough that the labels argued the underlying model must have trained on their members' actual recordings without a license. Udio never disclosed its training sources, which is itself a recurring flashpoint across the AI-music suits. With thousands of registered recordings across the three catalogs potentially at issue, statutory damages alone put maximum exposure in the billions before any accounting of profits or injunctive relief reaching into Udio's product itself.

From litigation to licensing

The case changed shape without a court ever reaching the central fair-use question. On October 29, 2025, UMG announced it had settled with Udio and, in the same announcement, unveiled a strategic licensing agreement for a new Udio platform built on licensed catalog, expected to launch in 2026. Warner Music Group followed on November 19, 2025, with its own settlement paired with a similar licensing partnership. Neither company disclosed financial terms.

That left Sony Music alone among the three majors, still pursuing its claims against Udio over the same underlying training conduct. Public reporting through spring 2026 describes Sony as the sole remaining plaintiff, with no fair-use ruling expected before 2027. Separately, the American Federation of Musicians has sued UMG and Warner over how their settlements treat working musicians — a distinct dispute that doesn't affect this docket, but underscores how contested the settlement terms are among the people whose recordings were actually used, including questions about how individual artists and songwriters opt into the new licensed platform and get paid once it launches.

What the settlements do not resolve

It's worth being precise about what happened here: two labels traded their claims for a business relationship, not for a court ruling that Udio's original training run was lawful. No judge has held that training a generative music model on unlicensed recordings is, or isn't, fair use in this case. The settlements resolve UMG's and Warner's exposure and buy them a stake in a licensed successor product; they say nothing about whether the pre-settlement training was infringing, and they do nothing to shield Udio from Sony's claims over that same conduct.

Where it stands

As of July 2026, the case is active as to Sony Music only. There is no trial date, no fair-use ruling, and no public indication Sony is close to a deal of its own. The parallel Suno litigation in Boston continues on a similar track.

Why it matters for training-data licensing

Udio is the clearest example yet of AI copyright litigation converting directly into a licensing market before a court decides the underlying legal question. For data buyers, the UMG and Warner deals establish a template: retroactive peace bundled with a forward license tied to a controlled, licensed platform. That structure is attractive precisely because it sidesteps the fair-use question rather than resolving it — and it lets both sides avoid the multi-year wait for an appellate answer.

For anyone diligencing a dataset or a music-generation product, the lesson runs the other way. A partial settlement doesn't clear the underlying corpus. The same training run that released Udio from UMG's and Warner's claims remains fully actionable by every rights holder who didn't sign — in this instance, one of the three biggest labels in the world. Chain of title on a training corpus isn't settled just because some of the plaintiffs went away, and buyers should ask specifically which catalogs are covered by any licensing deal a vendor points to, not just whether a lawsuit was resolved.

Buyers evaluating a music-AI vendor that advertises a settled lawsuit should ask three concrete questions: which specific rights holders actually settled, whether the deal covers the original training corpus or only a forward license for a new product, and whether any co-plaintiffs remain active against the same underlying conduct. In Udio's case, the honest answer to that last question is Sony Music — a reminder that a headline settlement can coexist with live, unresolved exposure over the exact same dataset.