How AI Tracks Are Hijacking Real Artist Pages and Diverting Royalties
Scammers are placing AI-generated songs on the official streaming pages of real artists and collecting the payouts themselves, according to an investigation first published by 404 Media on Sept. 17.
The proceeds from those fake tracks go to the uploaders, not to the artists whose names and pages carry them. That royalty diversion is the point of the operation. It turns impersonation into a payment rail. Engadget
The method is low cost and entirely built on legitimate tooling. The investigation describes a workflow in which an operator generates a track with Suno or Udio, generates cover art with an image model, then uploads the bundle through an open distributor such as DistroKid while falsely claiming to be an existing band. 404 Media tested the path against Brooklyn band Lathe of Heaven to confirm the loophole works end to end. Engadget
Once ingested, the fake does not sit in a parody account or a similarly named profile. It appears as a legitimate new release from the real artist on Spotify, Tidal, Apple Music, Amazon Music and other services. The money follows the upload. For listeners browsing a discography, there is no visible break in artist URI mapping or catalog continuity to signal fraud.
Spotify told 404 Media that profiles that are not actively managed can be especially vulnerable to this scam. That detail matters for infrastructure reasons. Distributors assert metadata, including artist identifiers, at upload time. Streaming back ends largely trust that assertion unless an artist, label, or rights holder disputes it through Spotify for Artists claims or takedown. Unclaimed profiles lack that monitoring loop.
Spotify's stated policy is that vocal impersonation is only allowed in music when the impersonated artist has authorized the usage. The policy was detailed in a September 2025 update. Spotify Enforcement, in practice, depends on detection and on someone with standing to flag the impersonation.
The scale is already beyond isolated incidents. The group Odette Child began tracking the problem in July and has catalogued more than 300 AI-generated songs impersonating acts including Taylor Swift, other celebrities and dead jazz musicians. Its work suggests operators are targeting both high-traffic names where even a small share of misattributed streams pays, and long-tail catalogs where no one is watching for new additions. Engadget
404 Media published its findings under the title 'I Hijacked a Real Artist's Spotify with AI Music. It Was Disturbingly Easy'. 404 Media
This is not the first collision between open ingestion and royalty attribution. In November 2024, AI-generated albums that did not belong to them appeared on the Spotify pages of Standards, Health, Annie and Swans, a case tied to distributor metadata failures. In July 2025, falsified work appeared on the pages of artists including Blaze Foley and Toto, prompting singer Leith Ross to publicly criticize AI music after manipulated songs surfaced under that name. The Verge Billboard
Earlier variants used human-made audio. Billboard reported in 2022 that operators hijacked viral hits by uploading their own copies of recordings, sometimes under the same artist name, to siphon royalties, and separately that uploaders gamed discovery playlists with tracks claiming to feature popular acts with no involvement. Billboard Spotify has also previously faced allegations that it gamed its own royalty system by creating and promoting in-house 'fake' artists, allegations it denied. The Verge
The broader context here is that streaming solved distribution and left identity under-specified. Anyone who has managed catalog ingestion knows the trade-off. Frictionless upload through aggregators enabled independent music at global scale, but it pushed verification to the edges. Distributors compete on speed and price, not on artist authentication. Platforms compete on catalog completeness, not on pre-release vetting.
In my view, that architecture will have to change without closing the openness that made it valuable. Stronger artist claiming, cryptographic linkage between distributor accounts and verified artist profiles, and anomaly detection for unexpected releases on dormant pages are all tractable engineering problems. They cost money and add friction, which is why they have lagged fraud. Generative audio simply lowered the cost of producing plausible filler to near zero, so the existing gap now scales.
Worth flagging for working technologists is where leverage sits. Detection at the audio layer will always trail generation. Detection at the identity and payments layer has a better chance. Who asserted this artist ID, from which distributor account, with what payment destination, and has that account ever delivered for this artist before. Those signals exist today in ingestion logs. Using them aggressively would not stop AI music, nor should it. It would make impersonation unprofitable, which leaves room for the legitimate uses to grow.
