An all-AI rock outfit calling itself Velvet Sundown had quietly amassed over one million monthly listeners on Spotify, and nobody in charge of verifying artists, streams, or press statements had caught on.
Industry executives first dismissed the numbers as an anomaly. Then they learned a self-described “privacy researcher,” not the band’s makers, had been fielding reporter calls. By the time editors realized they had quoted a phantom press agent, the episode had exposed cracks in every guardrail meant to protect artists and audiences from synthetic fakery.
An AI Band Climbs the Charts Unchecked
Velvet Sundown released two albums in rapid succession this spring; the two full-length albums amassed hundreds of thousands of streams within weeks of their release. Cover art depicted four retro-styled rockers, yet no such quartet exists. Spotify’s verification badge still appeared, and the band even landed on popular Spotify playlists dedicated to Vietnam-era classics.
The ruse gathered pace once Canadian technologist Andrew Frelon posed as a spokesperson. Between March and late June, the group’s audience jumped by about 700,000 listeners, fuelled by curious clicks and algorithmic recommendations. Frelon later admitted he had no affiliation and used ChatGPT to craft provocative social posts that baited coverage.
Detection Tools Give Mixed Scores
Streaming platforms trumpet new AI filters, yet results vary wildly. France-based Deezer says it keeps synthetic tracks away from its editorial lists and claims its in-house technology can identify Suno or Udio files with 100 percent accuracy. Overall, Deezer claims 98% detection accuracy.
Even so, the volume is daunting. Deezer now ingests more than 20,000 AI-generated tracks daily, a figure that has doubled since the beginning of the year. Synthetic uploads account for 18 percent of all new songs there, yet AI tracks still make up just 0.5 percent of Deezer streams. The company blocks suspect files from autoplay and editorial slots; Deezer excludes AI tracks from algorithmic playlists to protect payouts for human artists.
Fraud remains a stubborn problem. Deezer estimates that up to 70% of streams are fraudulent when a track is completely machine-generated. Spotify has disclosed no comparable metrics, and its discovery algorithms helped Velvet Sundown flourish undetected.
Fake Spokesperson Tricks the Music Press
Frelon set up fresh Gmail addresses and social handles, then pitched reporters with claims that record labels were trying to silence the “band.” Rolling Stone, Stereogum, and several radio outlets printed or aired his quotes. Editors failed to confirm his identity, and Spotify’s blue checkmark provided cover. By the time the guitar-free quartet’s real creators disowned him, the tale had become a hall-of-mirrors case study: a hoax inside a hoax.
How did the deception slip through? Newsrooms cite shrinking staff, rapid deadlines, and the difficulty of vetting digital sources.
Record Labels Strike Back in Court
While the Velvet Sundown saga unfolded, major labels sharpened a legal spear. The Recording Industry Association of America (RIAA) filed lawsuits against Suno and Udio in June 2024 on behalf of Universal, Sony, and Warner, accusing the start-ups of training models on protected songs without permission. The lawsuits allege AI outputs resemble hits by Queen, Abba, and Mariah Carey.
Suno’s chief executive dismissed the complaint, saying Suno’s CEO insists outputs are completely new rather than stitched-together copies. The defendants face demands for financial damages and injunctions that could shape how generators operate.
Europe Tightens the Copyright Screws
Lawmakers in Brussels have already moved. The EU AI Act was passed with 523 votes in March 2024. Among its sweeping clauses, the statute requires the disclosure of training data summaries and stipulates that these lists must respect European copyright. The law requires AI firms to comply with EU copyright rules worldwide, regardless of where they obtained the data. Advocates argue that the transparency provisions could reshape AI globally; developers warn that they could hinder open-source research.
Implementation has begun, although national agencies acknowledge that staffing and technical expertise will be in short supply for months. Meanwhile, labels plan to cite the EU law as further leverage in the Suno and Udio cases.
Platforms Wrestling With Standards
At present, no two streaming companies label AI songs the same way. Spotify offers voluntary tags, but none appeared on Velvet Sundown’s page until after press inquiries. Apple Music remains silent on its approach. Deezer trumpets its tech, yet it still must police tens of thousands of daily uploads.
Industry groups propose watermarks embedded during audio generation, but musicians and open-source advocates fear unintended censorship. “Artists deserve tools that identify fakes without blocking invention,” says civil-rights scholar Tiana Reed.
What Happens Next?
The Velvet Sundown caper may have started as an internet gag, yet its ripple effects look anything but funny. The story pulled back the curtain on brittle verification checks, yawning legal gaps, and a public hungry to believe a good yarn.
Can streaming giants rebuild trust? They will need shared metadata protocols, real-time fraud screens, and better cooperation with journalists. Legislators, for their part, seem poised to wave in stricter disclosure laws from Seattle to Seoul.
As for Frelon, he says he has left the stage. The software that gave birth to Velvet Sundown has not. Anyone with a laptop and a credit card can spin up a dozen more digital bands by sundown tomorrow—and maybe give them a slick talker to answer the phone.
Key Takeaways:
- An AI-only act, Velvet Sundown, drew more than one million Spotify listeners before being unmasked.
- A fake publicist exploited newsroom shortcuts, highlighting the weakness of source-checking.
- Deezer detects most synthetic tracks, but it still battles inflated streams and 20,000 new AI-generated uploads each day.
- Major labels are suing Suno and Udio, claiming the generators infringe on Queen, Abba, and Mariah Carey’s copyrights.
- Europe’s new AI Act requires developers to disclose their training data and adhere to EU copyright rules globally.
- Streaming platforms lack a common standard for flagging or filtering AI-generated music.