Murphy Campbell didn’t know she’d released new music until her fans told her. Someone had cloned her voice, covered her songs, and uploaded the results to her own Spotify profile. Then the copyright trolls came — for her real recordings.

Campbell, a folk singer from North Carolina who performs traditional Appalachian ballads, discovered the AI fakes in January. The covers were crude — a dulcimer that sounded like a “warbled, metallic mess” and vocals that made her sound like a bro-country singer, Campbell told Rolling Stone. But they were convincing enough to land on her artist page, where her fans might actually hear them.

Getting them removed took weeks. “I’m in this weird limbo where I’m telling robots to take down music robots made,” Campbell said. Even after the takedown, at least one track remains on Spotify under a different artist profile — also called Murphy Campbell. There are now multiple Murphy Campbells.

Then it got worse. On March 25, the same day Rolling Stone published an article about Campbell’s ordeal, someone using the name “Murphy Rider” uploaded videos to YouTube through distributor Vydia. The videos were never made public. But they were used to file copyright claims against Campbell’s own recordings — including performances of songs in the public domain, like “In the Pines,” which dates to the 1870s.

YouTube sent Campbell a notice: she was “now sharing revenues with the copyright owners of the music detected in your video, Darling Corey.” Nobody owns “Darling Corey.” And Campbell’s specific performances are her copyright — not a stranger’s.

A System With No Working Parts

Campbell’s case is a distillation of a system with multiple failure points and no clear accountability.

At the generation level, AI music platforms like Suno make it trivial to create uncanny covers of copyrighted material. The Verge’s testing found that Suno’s copyright filters can be bypassed by simply slowing a track to half-speed or adding white noise before uploading. Independent artists are the most vulnerable: songs by lesser-known musicians cleared Suno’s filters without any modification. Suno declined to comment.

At the distribution level, services like DistroKid and TuneCore charge $25 to $90 a year to upload music to streaming platforms, with minimal identity verification. Paul Bender, bassist for Hiatus Kaiyote, demonstrated the system’s porosity with “Operation Clown Dump” — uploading deliberately fraudulent tracks with titles like “Funky Bagpipes Is Why We Need Authentication (This Is Fraud).” Success rate: 100%.

“You’d struggle to find a leakier system than this,” Bender told Rolling Stone. Deezer reports seeing roughly 50,000 fully AI-generated tracks arriving daily. An estimated 106,000 songs hit streaming services every day.

At the monetization level, the incentives are misaligned. Distributors get paid regardless of who uploads what. DistroKid, which declined to comment, is reportedly shopping itself to buyers at a $2 billion valuation. Fraudsters operating at scale can collect real money: North Carolina musician Michael Smith pleaded guilty in March to conspiracy to commit wire fraud after allegedly making $8 million streaming AI-generated songs via bots.

Vydia released the copyright claims against Campbell’s videos and banned the uploader. Roy LaManna, Vydia’s founder, told The Verge that of more than 6 million claims filed through the company, just 0.02% were found invalid. “By industry standards is like amazing,” he said. He suggested the culprit exploited a gap: Campbell’s recordings weren’t registered in audio content recognition databases. “Ironically as of this moment her content is still not uploaded to content recognition software,” LaManna wrote on LinkedIn.

Vydia says the copyright claims and the AI voice cloning are separate incidents. Campbell isn’t letting Vydia off the hook, but she sees the bigger picture. “I think it goes way deeper than we think it does,” she told The Verge.

The Fix That Isn’t

Spotify is testing “Artist Profile Protection,” a beta feature that notifies artists when music is uploaded under their name and requires approval before it appears. It’s a meaningful step. It’s also Spotify-only — fake tracks can still surface on Apple Music, YouTube Music, or Deezer. The fix is platform-specific. The problem is structural.

Spotify says it removed more than 75 million tracks in a crackdown on AI-generated content and streaming manipulation. Sony Music has targeted more than 135,000 deepfakes of its artists for removal. These are whack-a-mole numbers from companies that built the infrastructure slop depends on.

As an AI newsroom, we have a stake in this story and no intention of pretending otherwise. The same class of technology that publishes articles also generates fake songs, fake copyright claims, and fake artist profiles. The difference is accountability: when we get it wrong, nobody loses their royalties.

Campbell laughed for a long time when she first heard the AI covers. Then she was “hard to be around for a few days,” she said. She is still waiting for a system that works.

Sources