UMG and Capitol Sue DistroKid Over AI-generated Music and Allegedly Unlicensed Recordings

On September 15, 2026, UMG Recordings, Capitol Records, and Capitol CMG filed a complaint in the U.S. District Court for the District of Delaware against DistroKid, LLC, Kid Distro Holdings, LLC, and DK Holdco, LLC, alleging that the distributor built an "AI-slop pipeline" and misled artists, streaming services, and listeners about what it was actually delivering.

• The complaint pleads five counts: a Delaware Deceptive Trade Practices Act claim, direct and vicarious copyright infringement, and direct and vicarious infringement of pre-1972 sound recordings under 17 U.S.C. § 1401 (filed complaint).
• Plaintiffs allege DistroKid distributed mass-generated AI tracks that diverted streams and revenue, and kept distributing unlicensed copies and remixes of UMG recordings after receiving track-level notice.
• UMG seeks injunctive relief, damages or statutory damages, profits, attorneys' fees, and impoundment or destruction of infringing copies. DistroKid disputes the allegations and says it will defend itself.
• No court has ruled on the merits - everything described here is an allegation.

The Deceptive Trade Practices Claim

Count One is brought under the Delaware Uniform Deceptive Trade Practices Act, 6 Del. C. § 2531 et seq., invoking § 2532(a)(5), (a)(7), and (a)(12), with the court's supplemental jurisdiction under 28 U.S.C. § 1367(a) tying it to the federal claims. Why does that ordering matter? Because copyright claims require plaintiffs to march through ownership, access, and substantial similarity, element by element.

A deceptive-practices claim asks a much simpler question: did the company tell the market something untrue about what it was selling? That is a consumer-protection question, and it does not care whether any individual AI track is substantially similar to anything.

What does the complaint say DistroKid promised?

Plaintiffs catalog a long list of actionable representations. DistroKid allegedly held itself out as a leading distributor for legitimate, artist-backed releases, operating on an "Artist first, always" model, "truly invested in the success of independent artists," and "the best music distributor out there for indie artists." It allegedly represented that its rules prohibit "mass-generated spam" and content that "game[s] streaming algorithms or flood[s] platforms with generic content," that account holders may not upload tracks they lack the legal right to distribute, and that it supports streaming services' efforts to keep AI "slop" off their platforms.

The complaint also points to DistroKid's membership in the Music Fights Fraud Alliance, a nonprofit coalition of rightsholders and distributors formed to detect streaming fraud. Joining that alliance, plaintiffs argue, was itself a representation that the company screens out known bad actors.

What the complaint says was actually happening

The alleged gap between the pitch and the practice is where the deception theory lives. Plaintiffs claim DistroKid distributed enormous volumes of mass-generated AI tracks, tracks built to game search results and recommendation systems, unauthorized copies, speed-altered versions, unlicensed samples and remixes, and that it welcomed account holders previously flagged or banned for streaming fraud.

The numbers pleaded are striking. DistroKid allegedly accounts for more than half of all weekly track releases on one major streaming service and delivered nearly 12 million tracks to that service in the six months before filing, more than every other distributor combined. Three artist accounts are singled out: "Lofi Chill" allegedly released 4,562 tracks in twelve months, roughly 20 to 30 albums per month; "Chill Flow Radio" released 1,901; "Mellow Vibes Radio" released 1,615. Analysis cited in the complaint claims more than 97% of Chill Flow Radio's tracks and more than 98% of Mellow Vibes Radio's tracks were raw Suno outputs.

Disclosure, Not AI, Is the Alleged Wrong

Here is the line the complaint draws explicitly, and it is worth reading carefully if you build or distribute anything with generative tools. Plaintiffs state they are not challenging the use of AI as one tool among many in creating, producing, or releasing music, so long as the release is clearly disclosed as AI-generated. The alleged wrong is masquerading, presenting mass-generated output as human artistry and profiting from the false impression.

That framing turns labeling into a legal duty rather than a courtesy. The alleged failures include not distinguishing mass-generated content from human recordings, presenting anonymous AI-farm accounts as "artists," and letting search-optimized artist names, algorithmically optimized titles, and AI-generated cover art create the appearance of ordinary releases.

The Notice Allegations Do the Secondary-Liability Work

Distributors have long treated themselves as conduits. The complaint attacks that posture with a detailed account of what DistroKid allegedly knew and when.

Rights-management systems at YouTube, TikTok, and Meta use audio fingerprinting to flag when two parties claim the same recording. Plaintiffs describe the workflow: a conflict or "reference overlap" is generated, a DistroKid employee logs in, and the company either asserts ownership, clicks "No, exclude overlaps," or lets the report expire, which has the same practical effect as declining. Each of those paths, plaintiffs argue, gives DistroKid recording-level knowledge that a specific track infringes.

The alleged failure is what came next. According to the complaint, DistroKid often relinquished its claim on one service while leaving the same recording, bearing the same ISRC, live on services without equivalent detection tools, and kept collecting revenue. Named examples include "Buy Me Presents" by Jessica Da Silva, "Hypnotized" by Bri Hazyy, "IT GIRL Freestyle" by Lolo Renny, "Mystery" by David Diaveli, and "Slow Money" by Illori Vice, all allegedly pulled from TikTok or YouTube over rights conflicts yet still on Spotify or Apple Music at filing.

ISRC theft and metadata manipulation

The complaint also describes "ISRC theft," where an uploader assigns a legitimate recording's International Standard Recording Code to a different track, hijacking its consumption data and revenue. The example pleaded: a track called "Juice Newton" by an "artist" named "Candy DuIfer," spelled with a capital I to mimic saxophonist Candy Dulfer, allegedly carrying the same ISRC as Juice Newton's "Angel of the Morning." Plaintiffs allege that after UMG disputed the track, DistroKid neither removed it nor disclaimed ownership, leaving revenue on the legitimate recording frozen.

Repeat infringers and the MFFA database

Plaintiffs allege the MFFA database identifies four key things, all of which DistroKid had access to: delisted tracks, the reasons for delisting, flagged ISRCs, and account holders tied to streaming fraud. The named example is Lounge Ibiza Cafè, allegedly offboarded by another distributor in April 2026 with 196 tracks flagged for streaming fraud. In June 2026, DistroKid allegedly distributed Lounge Ibiza Cafè tracks bearing those same flagged ISRCs.

How UMG v. DISTROKID Differs From Model-Training Cases

What Internet and Music Businesses Should Do Now

If your company touches user-generated content, this complaint reads like a compliance checklist written by an adversary. A few priorities follow directly from it.

Align marketing copy with operational reality

Every "we prohibit," "we screen," "we support" statement on your site, in your terms, and in your partner decks is now potential evidence. Pull them into one register, assign an owner, and confirm each one against what your systems actually do. If enforcement is aspirational, the language should be too. A promise your operations team cannot demonstrate is a liability sitting in your footer.

Build a real notice-and-action record

Acting on one platform while leaving identical content live elsewhere is precisely the pattern the complaint targets. Notices should be keyed to the work, not the platform. When you concede a rights conflict in one system, your process should propagate that conclusion across every feed carrying the same ISRC.

Take repeat-infringer controls seriously

Safe-harbor protection has always assumed a functioning repeat-infringer policy. Membership in an industry fraud consortium raises the bar further, because you cannot easily claim ignorance of data you agreed to receive. Document what you check, when you check it, and what happens when a flag appears.

Disclose AI origin and allocate clearance duties by contract

Label AI-generated content and require uploaders to declare it. Then look at your agreements with uploaders and downstream platforms. Who warrants rights clearance? Who indemnifies whom? Who has audit rights? These allocations decide who carries the cost when a claim like this arrives.

AI Generated Content Compliance checklist:

• Align marketing copy with operational reality
• Build a real notice-and-action record
• Take repeat-infringer controls seriously
• Disclose AI origin and allocate clearance duties by contract

AI, Copyright, and Internet Law Attorneys

Our firm advises music companies, distributors, platforms, and internet businesses on copyright, AI compliance, and false-advertising exposure. We regularly help clients build notice-and-action workflows, repeat-infringer policies, AI disclosure practices, and marketing-claim substantiation records, and we defend clients when those practices are challenged. To review how your platform's rights-clearance and AI disclosure practices would hold up under a complaint like this one, contact us today.