Why most AI takes are wrong.

The conversation about AI in music has collapsed into two camps. One says it is theft and anyone touching it is a fraud. The other says anyone skeptical is a Luddite who will get left behind. Both are loud. Both are missing the point.

I am not writing this to tell you AI is good or bad. I am writing this because the discourse has flattened a genuinely complicated conversation into a slap fight, and the people who are actually making music, running labels, managing artists, and building the next decade of this industry deserve something better than a slap fight.

Here is how I am actually thinking about it.

1. The training debate is not the whole debate.

The loudest argument against AI in music is about training data. Generative models learned from copyrighted recordings without licensing, without consent, and without payment to the artists whose work was the fuel. That is a real argument. It is being litigated in court right now. Suno settled with Warner. UMG and Sony are next. Labels are negotiating. Artists mostly are not in those rooms.

I am not going to dismiss that concern. It deserves the conversation it is getting.

But it is one conversation. Not the whole conversation. And once you notice that, you start to hear how many different arguments are being stuffed into the same sentence when someone says “AI in music.” Training data is one argument. Vocal cloning is another. Flooding streaming platforms with AI slop is a third. Using AI as a production tool in an otherwise human workflow is a fourth. These are four different problems with four different answers. Treating them as one is how the conversation broke.

2. Both sides have logic. Neither has nuance.

The case against AI in music is real. Artists’ work was used without consent. Platforms are being flooded. Deepfakes are a problem. Session singers, engineers, and producers are watching a version of their job get automated in real time. None of that is hysterical. All of it is happening.

The case for AI in music is also real. The tools are lowering the cost of making music to near zero. A kid with a laptop can now access what used to cost a studio booking and a producer’s week. Not every AI use is generation. Transcription, stem separation, mixing assistants, research, admin, scheduling. Most of what AI is doing in music right now is not making songs. It is removing the friction between an artist and their own ideas.

Both sides have a point. What neither side has is the patience to hold both points at the same time. That is the conversation I want to have.

3. “AI in music” is not one thing.

This is the most important thing I will write in this piece.

When someone says “AI in music,” they could mean any of the following:

• A model trained on your catalog without your permission.
• A deepfake of your voice singing a song you did not write.
• A Spotify account uploading 400 fake tracks a month to farm royalties.
• A producer using Suno to sketch a direction for a real record.
• A manager using Claude to draft a press bio in 10 minutes instead of 10 days.
• An indie artist using stem separation to rework a sample they cleared.
• A bedroom artist using an AI mixing tool to finish a song they otherwise could not afford to finish.
• A manager or indie label building direct-to-fan infrastructure, sites, CRM flows, email campaigns, merch drops, without hiring a full engineering team.

That last one is the one nobody is talking about. Building direct-to-fan has never been easier. AI is letting a two-person team do the work that used to require a five-person team and tens of thousands of dollars in engineering cost. That is not a theoretical benefit. That is happening right now, at every indie operation I know.

These are not the same thing. Some of them are theft. Some of them are tools. Some of them are scams. Some of them are access. When you argue about “AI in music” without specifying which one, you are not having a debate. You are having a vibe.

If you take one thing from this piece: next time someone tells you how they feel about AI in music, ask them which one.

4. Diplo said the quiet part.

“You’re not going to win. There’s no fighting AI. You have to just work your best to be the best at it right now. You’re just wasting a year being like ‘ahh’ because everyone else is going to use it and not give a fuck what you think.”

Diplo, on the Behind The Wall podcast, April 2026.

He took heat for this. Most of the heat missed the point he was making.

Diplo is not saying AI is good. He is saying it is inevitable. There is a difference. When a technology drops the cost of a thing by 95%, that thing changes, whether you approve of it or not. You can decide how you personally want to engage with it. You cannot decide whether it exists.

The real question is not whether you like it. The real question is what you do now that it is here.

5. The floor just fell.

Here is the story nobody is telling.

For the last 30 years, making a professional record required a studio, an engineer, a producer, and enough money to book all three for as long as the record took to make. For most of that history, that meant a label. A label meant a contract. A contract meant giving up control of your masters, your publishing, your schedule, and in a lot of cases your creative direction.

The bedroom-studio revolution of the 2010s cracked that. Billie Eilish and Finneas made an album that swept the Grammys on a $3,000 setup in a childhood bedroom. That was the first big crack.

AI is the second crack, and it is a much bigger one.

A teenage artist in 2026 has access to tools that, 10 years ago, cost a six-figure studio build. Vocal production. Mastering. Arrangement. Research. Pitch decks. Scheduling. Distribution. The floor on who gets to try this just fell through. Not because AI is doing the work. Because AI is removing the stuff that used to sit between the artist and the work.

The same is true for managers, indie labels, and independent operators. People who used to need a team of five to run a release campaign can now run it with two and a stack of tools. The friction that used to protect the incumbents is going away.

That is the story I wish more people were telling. The music industry spent 20 years gatekeeping access, and a lot of the current panic about AI is panic about losing the moat.

6. The grift is real.

I do not want to pretend the concerns are imaginary. They are not.

Xania Monet debuted on the Billboard R&B radio chart as the first “AI artist” to do so. Real songwriter behind the scenes, AI doing the music and the voice, brand-new avatar. Spotify is being flooded by accounts uploading hundreds of AI-generated tracks a month to farm royalties from the playlist economy. Voice clones are everywhere. Some of them are used for fun. Some of them are used to release fake tracks that get millions of streams before they are pulled down.

This is happening. It is bad. And it is going to get worse before it gets better.

But here is the trap: if you let the worst examples define the whole category, you will walk away from a technology that is also making it easier for the artist you actually want to find to finish their first song. Fakers will fake whether AI exists or not. They were faking with bot farms, payola, fake followers, and paid-for press long before generative models existed. The grift is not new. The tool is.

Do not let the worst users of a thing decide how you think about the thing.

7. Finneas asked the right question.

I want to end on the question I think about the most. Finneas said this recently:

“If I were broke again and 17 again, I would be figuring them out. I would be figuring out how to get something juicy out of them. Because it’s free. So I don’t want to poo-poo a thing that I think is accessible to everybody. If it inspires you, I think that’s cool. But then, do you feel like you made it?”

Finneas, on AI in music.

That is the question. Not “is AI theft.” Not “will AI replace artists.” Not “should we ban it.”

Do you feel like you made it.

Because that is the test. Every artist gets to answer it for themselves. For some of them, using Suno to sketch a scratch demo is not making music, it is shopping. For others, the AI is a brush, and what comes out the other side still feels like theirs. The tool does not answer the question. The artist does.

What I would not do is answer it for anybody else. And what I would not do is let people who are not making anything tell people who are making something how to feel about their own process.

8. Don’t let the grift define the tool.

The case I am making in this piece, in one line: AI in music is going to be abused, is already being abused, and none of that is a reason to walk away from it.

The grifters are loud. The takes are dumb. The discourse is broken. But the artists who are going to define the next 10 years of this industry are not sitting around arguing about it. They are figuring out which tools help them make more of what they actually want to make, and which ones get in the way.

That is the shortlist worth paying attention to.