The Velvet Sundown: The AI Band Controversy Explained

Everybody is talking about the Velvet Sundown, a prolific band that, in just a few months, released three albums and garnered over 1.4 million monthly listeners via its verified Spotify account.

If that sounds questionable, it is. The band has recently confirmed that its members are "not quite human, not quite machine" on its official social media accounts. 

Who Are the Velvet Sundown? A Timeline

The Velvet Sundown is a band comprised of, according to Euronews, vocalist Gabe Farrow, guitarist Lennie West, bassist-synth artist Milo Rains, and percussionist Orion "Rio" Del Mar. In June 2025, the group debuted two albums, Dust and Silence and Floating on Echoes on Spotify. Reddit users started to question the band's legitimacy after finding the Velvet Sundown's songs populating their Discover Weekly playlists and not being able to find much about the group online.

In late June, after publications started to cover the band, a person under the pseudonym Andrew Frelon started a fake X account and pretended to be a spokesperson for the Velvet Sundown in what turned out to be an elaborate media hoax. Frelon fooled a lot of folks, including Rolling Stone. 

On July 3, the "real" band confirmed on their official social media accounts that they are a "synthetic music project guided by human creative direction, and composed, voiced, and visualized with the support of artificial intelligence."

"This isn’t a trick," they add, "it’s a mirror. An ongoing artistic provocation designed to challenge the boundaries of authorship, identity, and the future of music itself in the age of AI." 

On the same day, Frelon published a Medium post explaining why he was posing as the band. It says that the hoax exposed "numerous and large gaps in the verification process used by the majority of journalists, most likely occurring especially when they are chasing after a large and timely story such as that surrounding TVS. (Also, to be fair to them, I exploited a confusing situation around account ownership based on my professional knowledge in this domain. But they still should have done better.)"

Euronews posted a statement from a Spotify spokesperson, noting that "we don't prioritize or benefit financially from music created using AI tools. All tracks are created, owned, and uploaded by licensed third parties."

And the band is still creating; a third album, Paper Sun Rebellion, dropped on July 14. As of this writing, their most popular song, "Dust on the Wind," a ’70s Southern rock-inspired tune, has nearly 2 million streams. 

We chatted with Berklee faculty about the Velvet Sundown and the rise of AI music. Here's what you need to know:

AI Music and the Rise of Ghost Artists

"Ghost artists have been part of music culture for generations, and AI is only the newest tool in that lineage," says Ben Camp, associate professor of songwriting. 

"Modern listeners already cheer for artists who hide their faces or even their human form. Gorillaz present themselves as animated characters. Daft Punk performed behind helmets. MF DOOM rhymed in a mask. Hatsune Miku is a vocaloid hologram who draws stadium crowds that sing along to all of her hits. None of these acts were produced by AI, but they all maintained some level of anonymity for the artists. The anonymity was even a point of engagement for some fans, sparking debate or speculation about who the band is. Regarding the Velvet Sundown, I was surprised by how much the discussion was about the artist’s identity, rather than the music, lyrics, and performance."

"On the other hand, the talented Martha Wash sang 'Everybody Dance Now,' but the band C+C Music Factory kept her from the public. Milli Vanilli was panned for singing to a pre-recorded backing track. Anonymity is a choice, and it comes with both benefits and drawbacks," they add.

Songwriters are often already anonymous, with Camp mentioning that many artists don't write their own songs. They cite Beyoncé’s “American Requiem,” which credits 15 writers and producers.

AI technology helps makes the process of music making more accessible, they add. "Fewer people are excluded from creating because . . . they grew up in a small town without a music scene, or because a disability is preventing them from performing their music."

How to Spot AI-Generated Music

There are ways to tell if music is AI-generated, although the signs shift quickly, Camp says. "Musically, it is becoming difficult to know whether a melody or chord progression started with a human or a model. Right now, sonic quality is the clearest clue. Some AI tracks still feel slightly compressed or 'watery,' though fidelity improves month by month. A few years ago, the output was unusable in a pro setting; now it often is."

In terms of lyrics, Camp says AI might be a little more simplistic than human creations, but not always. "Early AI work can lean on strict rhyming couplets, uniform line lengths, or a hokey greeting‑card style tone. But those traits also show up with first‑semester songwriting students until they really learn their craft. Whether you collaborate with a person or a machine, it's worth it to craft and refine the song until it carries your own voice."

Similarly to samplers or synthesizers before it, AI is a tool, not a genre.

Ben Camp, associate professor, songwriting

It's only going to get harder to tell the difference. "At the current pace of improvement, in a few years time I'd be surprised if I could reliably distinguish 'AI Generated,' 'AI Assisted,' and 'No AI at All.'"

Camp says it's a blurry line between what is "human music" or "AI music" as it is. "I have many students who start with their own lyric concept, iterate it with an LLM [large language model], generate some verses and pick a few lines they like, add their own touches, and go back-and-forth until they have a finished product. Is that AI music? Or human music?"

"I know musicians who will hum a riff, feed it into AI, generate a full arrangement around it, get the stems, bring it into their DAW, sweeten up the mix, and add a few instruments on top. Is that AI music, or human music? Similarly to samplers or synthesizers before it, AI is a tool, not a genre." 

What Are Streaming Services' Policies on AI-Generated Content?

Currently, Spotify's policy, per Descript, allows for AI tools in the creation of content; it does prohibit AI music that impersonates existing artists, as well as using content on its platform to train AI models. Descript also notes YouTube and Audible's current policies, which continue to evolve. 

One legal issue is that AI models are often trained using art without artists' consent. "While on the surface, the current copyright code is sufficient to address works created by AI (and thus provide legal protection and recourse for artists), all stakeholders who believe in and desire a sustainable music industry ecosystem must work aggressively to ensure that the copyright code is enforced and artists must be educated with respect to their rights," says George Howard, professor of music/business management. "The alternative is nothing short of the continued eradication of the chance of sustainability for the vast majority of artists."

"The music industry is a canary in the coal mine for pretty much everything, and so it is both not surprising that music is at the fore in terms of this technological shift, and it will not be surprising when it reaches all other segments of our lives," he adds.

"There can and will be certain outcomes that are net positive for artists via this new technology, but make no mistake, what we are confronting with respect to AI and soon AGI [artificial general intelligence, where machine thought is equal to or surpasses human capabilities] is a far different challenge than any other historical technological innovation and one that likely will, unlike others that were ultimately neutral to positive sum for participants, be zero sum."