When Someone Tells You AI Music Is Inevitable, It’s OK to Mock Them.

Last week, Bandcamp announced a ban on music made wholly or largely with AI, a move the musician Holly Herndon lambasted. Why should we submit to something else we don’t want?

Welcome to the eighth edition of Out & Back, my new newsletter about music and whatever else. If you haven’t subscribed yet, I’d be honored if you did. And if you like what you read, pass it to a friend. This week’s edition is about Bandcamp, AI music, and opinions that are a drag. Thanks for being here.

Early last week, a friend sent me a song he wanted me to hear. It was a mid-tempo soul ballad about love as a balm, its lyrics about “drifting into your tide” and “falling with you” calmly rendered by a singer who had little urgency and, for that matter, even less dynamic range. I didn’t hate the song; it felt like walking into a pleasantly lit room, with a stick of subtle incense burning in an unseen corner. It was just so anodyne—the lyrical treacle, the drums that felt like they were digging into someone else’s pockets, the guitar as soggy as day-old spaghetti—that I never needed to hear it again.

My pal wasn’t sending it over, though, for a quality assessment; he wanted to know if I could tell it was probably the work of AI. The song was, as you might have surmised, “Into the Blue,” one of three dusky tunes by Sienna Rose, a singer nearing four million monthly Spotify listeners and sporting the dubious “Verified Artist” blue badge on the platform. Except, as the BBC reports, “All the signs indicate she’s not real,” or that she is the product of a programmer with an affinity for soft-lidded soul. Her grainy, filtered portraits portray her as a classic soul singer, her curls resting on a pillow or mouth opening to an old Unidyne microphone. She is, at least, some machine’s projection of that idea.

If you haven’t had one of those moments yet, where a friend sends you a song to see if it passes your machine test, you surely will soon. Last November, the streaming platform Deezer reported it received more than 50,000 AI-generated tracks every day, up from 30,000 only two months earlier. And whether it’s horny country or sappy Sienna Rose love songs, three of which sailed smoothly into Spotify’s Viral Top 50, the stuff is seeping into or near our daily listening lives. (I use “the stuff” here because I think the term “AI music” is nonsense, relinquishing the art-making privilege that makes us human. I am sure I will use it in a bit, but I resent the phrase, anyway.)

Karl Kuehn, the artist behind this great shirt (Photo by Nick O’Reilly)

Within the next year, I suspect the stuff won’t be passed around as a novelty; it will simply be passed around as music. The prospect of AI music, after all, eliminates the limiting factors of making something to sell—human toil, fatigue, imagination. As more corporations realize they can cut expenses by cutting out the human creator, that is what they will do. AI music, then, is an economic inevitability; it does not, however, have to be an artistic inevitability. Anyone who tells you otherwise is a bootlicker.

This conflict has been on my mind more than normal in the dawning days of 2026, and not just because of Sienna Rose. The same day I heard “Into the Blue,” Bandcamp announced it was banning “music and audio that is generated wholly or in substantial part by AI.” Six months earlier, Deezer committed to tagging AI-generated sound as such and not slipping those songs into its recommendations. Bandcamp’s move, though, is a stark new line in the sand, an all-out attempt to prohibit something that would likely turn an increasing profit.

The decision raises significant questions, of course, yet to be answered—about a community self-policing itself, about resources required to detect increasingly sophisticated AI music, about what “substantial part” actually means, about where the continually shifting boundaries between AI and digital production begin and end. But it mostly reflects a fundamental belief that music is a human endeavor, that something isn’t art simply because it sounds like something a person might make but didn’t. “Music is much more than a product to be consumed,” Bandcamp said. “It’s the result of a human cultural dialog stretching back before the written word.” I’d call that true but melodramatic, if the whole clusterfuck didn’t seem so existential. It does.

I wasn’t exactly surprised to see the artist Holly Herndon push back on Bandcamp. For a dozen years now, she and husband Mat Dryhurst have been asking hard and important questions about how the rising tide of AI can work alongside the sinking state of artists. They have built Holly+, a “vocal deepfake” that turns other sounds into Herndon’s voice and was built in such a way that any commercial use cuts Herndon in for half of the profits. AI systems and models have become a substantial part of the pair’s shared artistic practice, like building an AI choral model by inviting 15 British choirs to sing a shared songbook and a video installation about the traumatic birth of their son, Link. As provocative as it often is, Herndon’s work will almost certainly fall inside of Bandcamp’s embargo and demand answers to some of those aforementioned questions.

But I was surprised by the myopic specifics of Herndon’s response. In a Twitter post (sorry, not calling it X) where she called Bandcamp’s strategy “a tourniquet,” she insisted that the “human / AI binary is not going to hold.”

Our choice to break that binary, especially in art, is exactly that—a choice. It is entirely possible to choose not to use artificial intelligence in your music, to create an artistic practice that inverts or at least elides Herndon’s own approach. It is possible to choose not to hear music made with AI, too, so long as we can still detect it. The binary holds every time someone decides, for themselves, to hold it. Herndon invoked another absurd binary in her retort, saying that she had “more authorship in [her] models than most pop stars do in their songs.” Most music is not the music of pop stars, and I don’t think that’s the ground Bandcamp is tending here. Pop music will probably go where the money goes; the rest of the music world does not have to follow.

I grimace every time I hear someone say that we, as a species or a society, have designed a system so strong and vast that we can no longer control it. Imagine the arcade of horrors that would still exist had we decided to simply abide the worst consequences of every system that humans had tried. It’s fitting that Herndon offered that pronouncement on Twitter, too; so many people who detest its owner’s actions and ideals still use it, because maybe it’s still fun or because they think their presence is too small there to matter. It’s almost like we’ve forgotten that individual action begets collective action, that it’s a vector of real change.

Herndon talks a lot about “infinite art,” the phrase that became the title of Anna Wiener’s smart and gently skeptical 2023 The New Yorker profile of her. “We live with infinite media,” she ended her tweet. She is referring to the endless possible outcomes of generative AI, how a system without the limits of humans—fatigue, hunger, a dearth of “ideas”—can simply keep making stuff.

But Herndon seems wildly dismissive of the finite actual art that enables those systems to exist at all. I recoiled at a line in a 2024 interview about the choir model that she and Dryhurst were building. “That’s something we’re focusing on for the show—showing the training data as art-making itself, as human-generated art-making itself,” she said of the people she’d asked to sing. What the fuck else did anyone think a bunch of human beings singing was aside from “human-generated art-making,” a phrase that gives a little corner of my brain a cramp?

Herndon likewise spoke in The New Yorker of her voice as “input,” like a wire plugged into a laptop. That’s fine for her, sure, but, for others, that is the apogee of their humanity, the output of their art. Something needn’t help build a model to be “art.” (Hilariously, one of Herndon’s partners in the company Spawning called other people’s artwork “data” in The New Yorker, then later told Wiener he tries to avoid that term. Totally, dude.)

AI is being foisted upon us, no matter whether you want it or not. It’s showing up as tools in our inboxes, slop in our social media feeds, propaganda in our politics. This year, data centers in the United States could consume the same amount of power as 40% of all American homes, and customers are paying the exorbitant costs for their construction in their utility bills. Again, it doesn’t matter if they want or reject AI. So can we not withhold our consent when it comes to the art we make or consume? That is essentially what Bandcamp has done as a company.

Maybe Herndon is right, that the move is nothing more than a short-lived tourniquet. But not trying is simply to bend to the relentless tide of capitalism, or to see the boot and lick it because you thought nothing else could or should be done. I fully reject that worldview, and I’m glad at least one music platform does, too. It is easy and perhaps tempting to look at AI discourse and joke that pantywaists had the same anxieties about the Internet or the automobile or the electric toothbrush; those technologies did not, however, pose an existential threat to the world we can make through art. I don’t think this makes me a Luddite. I think it makes me someone who is invested in humanity.

Last week felt like a long one. I was on four airplanes for two reporting trips, and the news in this country remains, of course, heinous. I was very tired by Friday night. But a few friends had invited us to a concert in a 140-year-old, one-room schoolhouse in a mining town cloistered deep in a Colorado canyon. It was an old-time show, a duo from a New Zealand fiddler named George Jackson and a Maryland banjo player named Brad Kolodner. Just before it started, I asked a friend a question about the year Colorado joined the Union. When he went to Google it, he saw that he had no cell phone service, no WiFi.

George Jackson + Brad Kolodner, last weekend in Salina, CO

For the next two hours, everyone in the schoolhouse sat silently and without screens, simply listening to songs older than the building and laughing at corny jokes about string bands. At one point, Jackson introduced a standard called “Billy in the Lowground” but explained that they would be playing a version called “Billy Lost in the Lowland,” recorded by the Kentucky powerhouse fiddler Darley Fulks. Oral tradition had warped the tune by the time it reached Fulks in his old Kentucky home, and this was the weird way he had done it.

I couldn’t help but think about Herndon, who grew up singing in church in Johnson City, Tenn., a place not that far or really that different from where Fulks had lived. I assumed she knew her home was one cradle of American music, the songs that poured out of it a century ago warped by the same oral tradition that shaped “Billy Lost in the Lowland.” They were all products of collective human imagination, the results of lineages as long as history itself, not of machines shaped by our history just to make someone else’s future more profitable.

(If you want to buy Karl’s great shirt, which I own myself, you can get it here.)

One more time, with feeling.