Suno AI signed, and the outrage sounds just like it did with Auto-Tune

In June 2024 the major record labels sued the AI company Suno. In November 2025 Warner Music signed a deal with that very company. In July 2026 a Munich court banned the same service from using "Atemlos" and "Daddy Cool" for training without a licence. In August 2026 the next licensing deal followed.

That sounds like chaos. It is. And underneath runs a debate I've heard before, just in different words: is it still music if a machine is involved?

My mother had this debate when I was a child. It was about Auto-Tune.

Before I have an opinion, the facts, each with a source. If you only catch the uproar, you'll think this is all one story. It's at least four.

• End of November 2025: Warner Music and Suno announce a partnership and end their legal dispute. Suno is to release new, licensed models in 2026 and switch off the old ones. Artists are to decide for themselves whether their name, image and voice may be used. No settlement amount is given in the announcement. (Warner Music, Rolling Stone)
• 31 July 2026: Munich Regional Court I rules largely in favour of GEMA, the German music rights society, in its case against Suno. Six songs are at issue, including "Atemlos", "Daddy Cool" and "Forever Young". The court finds that training with protected works is already a use that requires a licence. (GEMA, WBS Legal)
• 12 August 2026: BMG signs a licensing deal with Suno, covering recordings and publishing. Here too, artists have to actively opt in, and past use is settled as part of the deal. (Billboard, Variety)
• At the same time: Universal and Sony are still suing. Not everyone is on board, and whoever signed didn't sign on behalf of the others. (Billboard)

In short: the industry is suing over and licensing the same technology at the same time. That's not a contradiction, it's negotiation. Whoever sues is negotiating the price.

When I was little, there was always music playing at home, and my mother had a firm opinion about it: Auto-Tune and all that electronic music, those weren't real artists. No talent. Couldn't even sing properly.

I disagreed back then, and we argued about it a lot. Today I know: she was partly right. There were people who would never have made it onto a stage without the technology, and you could hear it.

And she was partly wrong, in the more important place. Music came out of exactly this technology that hadn't existed before. In 1998 Cher overruled her label to keep the robotic sound on her voice. The song became the biggest hit of her career, and for years afterwards the effect was simply called "the Cher effect". (NPR)

Eleven years later Jay-Z released a single called "D.O.A. (Death of Auto-Tune)". (Wikipedia)) In 2010 TIME magazine put Auto-Tune on a list of the fifty worst inventions of all time. The reason: a technology that makes bad singers sound good and truly bad ones sound like robots. (TIME)

Auto-Tune didn't die. Today it's in almost every production, usually so subtle that you no longer hear it. The argument disappeared; the tool stayed.

My claim: take away the dates and none of these uproars can be told apart from the others. And they're much older than most people think. Six events, every clue to the date removed.

Now the part that hurts, and I'll say it anyway.

The outrage runs under the headline: human versus machine. That sounds good. It's just that the image of the lone artist who does everything themselves simply doesn't hold for a large part of the charts.

A single songwriter, Max Martin, has written or co-written more than thirty number-one singles on the Billboard Hot 100. That puts him second of all time, behind Paul McCartney and ahead of John Lennon. (Wikipedia, Billboard) These songs were sung by a great many different people. Their faces are on the covers.

That's no scandal. Division of labour has been normal in music for a hundred years, and a good performer is an achievement in their own right. But it breaks one thing: the argument that pure human authorship is standing up to the soulless machine here. Authorship has been shared out for a long time. It was before AI existed.

So when someone says AI music is worthless because the performer isn't the author, they're defending an image the industry itself stopped delivering decades ago.

That brings us to the real question, and I think most people argue about tools because the real question is uncomfortable.

Vote before you read on. You'll see the result straight away.

My answer: the idea weighs the most, by a clear margin. Anyone who sits down with something in their head that doesn't exist yet, and wrestles with a tool until it's there, has created something. Whether the tool was a guitar, a sampler or a language model changes little.

And the opposite is just as true. Anyone who taps a machine three times and takes the first thing it spits out hasn't created anything. That isn't a work then, it's a lucky find. The difference isn't in the tool; it's in the intent and in the number of attempts.

Now I have to correct myself, because the comparison with Auto-Tune has a gap, and anyone who overlooks it is making it too easy for themselves.

Auto-Tune didn't take anything away from anyone. The software processed a voice that was there in the room. A language model for music comes about differently: it learns from recordings other people made, and without those recordings it can't do anything at all.

That's exactly what the Munich case was about. The court didn't rule that AI music is evil. It ruled that training with protected works is a use that requires a licence. (GEMA) That's a very sober statement, and it gets to the heart of it: the tool isn't the problem; the question is whether the people whose work is inside it get paid.

That's why I think the deals are the more important part of the news. A model that runs on licensed material, where artists have to actively agree, solves the problem. Outrage about machines being involved solves nothing at all.

In February 2025 more than a thousand musicians released an album of almost nothing but silence, recorded in empty studios, against proposed copyright rules in the UK. (NPR) That wasn't a protest against technology. It was a protest against being expropriated. The difference is everything.

I build websites, brands, images and animations, and the same debate has been going on in my craft for three years. Here's what I've taken from it:

• The tool decides nothing. There was never good work because someone had the expensive brush.
• Intent decides everything. If you know what you want, a language model makes you faster. If you don't, it just gets you to something arbitrary faster.
• The craft doesn't disappear, it shifts. The work used to be in the execution. Today it's in choosing, discarding and refining. That's more tiring than it sounds.
• Provenance becomes value. The more arbitrary stuff gets made, the more it counts that someone with a name stands behind it.

There will be a wave of bad AI music. A lot of it. Just as there was a wave of bad Auto-Tune music and a wave of bad websites after the first site builder.

And there will be a few people who make something with this tool that wouldn't exist without it. In five years we'll know their names, and then nobody will talk about what it was made with anymore. They'll talk about it being good.

The question isn't whether machines may be involved. The question is whether the people whose work is inside the machine get paid, and whether in the end someone actually wanted something.

My mother would disagree with me. She already did back then.

• Warner Music Group and Suno, announcement of 25 November 2025
• Rolling Stone on the settlement between Warner and Suno
• GEMA on the ruling of Munich Regional Court I of 31 July 2026
• WBS Legal with a legal analysis of the ruling (German)
• Billboard on the licensing deal between Suno and BMG
• Billboard on what the deals mean for the future
• NPR on twenty years of "Believe" and Auto-Tune
• TIME, Auto-Tune on the list of the fifty worst inventions
• NPR on the silent album by more than a thousand musicians
• Wikipedia on Max Martin

As of 19 September 2026. If any of this changes, I'll add it here.