50 MILLION SONGS WALK INTO A BAR. NOBODY WROTE THEM.

Spotify says it pulled more than 50 million slop and spam tracks from recommendations this year. Meanwhile, the music industry is figuring out how to make AI pay. Somewhere in between is the question nobody has answered: what happens when making music no longer requires a musician? Elena Stevens

Editorial Director, Riot + Reverie

Spotify says it has pushed more than 50 million “slop and spam” tracks out of its recommendations this year. At almost the same time, the company is moving forward with licensed AI-powered tools that will let fans create remixes and covers of existing songs.

So apparently AI music is both a problem and a product.

Welcome to 2026.

The argument over artificial intelligence and music has mostly been reduced to the obvious question: Should AI be in music at all? But that question is becoming increasingly useless. AI is already here. Artists are using it. Streaming platforms are dealing with it. Labels are licensing around it. Musicians are fighting over how their work can be used by it. The more interesting question is what happens when a technology that can help people make music becomes capable of making music without much help from people at all.

Music has survived plenty of technological panic before. Synthesizers were going to replace musicians. Drum machines weren't real drums. Sampling was stealing. Auto-Tune was cheating. Pro Tools ruined recording. Bedroom producers weren't “real” producers. Most of those arguments look ridiculous in retrospect because musicians did what musicians always do when somebody hands them a new machine: they fucked around with it and made something new.

AI doesn't automatically break that tradition. An artist using it to separate stems, restore a recording, experiment with an arrangement or create something they couldn't otherwise make isn't necessarily doing anything fundamentally different from generations of musicians who adopted whatever new technology showed up.

The problem begins when the tool stops expanding what the musician can do and starts making the musician optional.

WHEN THE TOOL BECOMES THE WORKER

That 50 million figure matters because generative AI changes something more fundamental than sound. It changes scale.

A musician is limited by time, money, energy and the irritating biological requirement to occasionally sleep. A generative system isn't. It can produce material at a speed that has almost nothing to do with human creative capacity, pouring more music into an ecosystem that was already carrying more songs than any of us could hear in a lifetime.

A person with a drum machine can make a record. A machine capable of generating music endlessly can make inventory.

That distinction matters.

The concern isn't simply that a computer can produce something resembling a song. Machines have been involved in making music for decades. The concern is what happens when songwriting, performance and production can increasingly be collapsed into a process requiring less and less human participation.

And yet, drawing a clean line between “human” and “AI” gets stupid pretty quickly. A songwriter might use AI-generated drums. A producer might use it to test arrangements. Someone physically unable to play an instrument might finally be able to realize music that has existed in their head for years. Another person might type a prompt, generate a finished track and upload it before lunch.

Trying to determine exactly what percentage of human involvement makes one legitimate and the other fraudulent misses the larger point.

The question isn't how much AI touched the song.

It's what happened to the human being on the other side of it.

FOLLOW THE MONEY

This is where Spotify's position gets interesting.

The company says AI-generated slop and spam don't belong in its recommendations. But Spotify is also developing a licensed system for AI-powered fan remixes and covers. A new agreement with Believe will allow participating Believe and TuneCore artists and labels to opt in. When those remixes are streamed, the artists and rights holders behind the original music will receive royalties.

In other words, Spotify isn't drawing the line at AI.

It's drawing the line somewhere around permission, control and compensation.

That's a much more useful conversation.

The American Federation of Musicians is having its own version of it. Its newly launched Human Made, Human Paid campaign is pushing for stronger AI protections in negotiations with the major labels, including the ability for musicians to approve or refuse the use of their work for AI training, along with transparency, credit and compensation. The union is also fighting Universal and Warner in court, alleging that recordings featuring its members were licensed to AI companies without properly compensating those musicians.

Strip away the futuristic language and suddenly this looks remarkably familiar.

Who owns the work? Who gets to use it? Who gets paid when somebody discovers a new way to make money from it?

The technology changes. The music business somehow keeps finding its way back to the same argument.

But money isn't the only thing at stake.

We already live inside an internet drowning in content. Artists are told to post constantly, release constantly, feed the algorithm, stay visible and never disappear. Now we've invented technology capable of producing essentially infinite material for an ecosystem that already has too much of it.

From a business perspective, the appeal is obvious.

From a human perspective, it's a little fucking bleak.

Music has never mattered because there wasn't enough of it. It matters because every once in a while somebody makes something that crosses whatever invisible distance exists between two people and says, you too?

Music isn't valuable because it's difficult to make, either. Plenty of difficult music is terrible. Some extraordinary music has been made with three chords, broken equipment and people who barely knew what they were doing.

The point is that somebody meant it.

Use the laptop. Use the synthesizer. Use the drum machine. Use Auto-Tune. Use AI. Break all of it. Make something weird. Technology helping a human being realize an idea is practically the history of recorded music.

Technology replacing the need for the human being to have an idea in the first place is something else.

SO, SHOULD AI BE IN MUSIC?

Probably.

That's what makes the argument uncomfortable.

The questions that matter now are harder: Who consented? Who gets credited? Who gets paid? And underneath all of them, did somebody actually have something to say?

If AI becomes another instrument musicians can bend, abuse and reinvent to make something we haven't heard before, music will do what music has always done.

It'll mutate. Good.

But if the destination is infinite songs generated for infinite playlists, optimized for infinite engagement, then we're not really talking about music anymore. We're talking about content. And Jesus Christ, we've already got enough of that.

Maybe the question isn't whether AI belongs in music.

Maybe it's whether music still needs us.