AI music label contract — what to check before signing

By Serj · August 2026 · 9 min read

Before we start: I'm not a lawyer, and nothing in this article is legal advice. This is a personal account of what I found when I read a real contract offer. Before signing anything involving your creative work and money, show it to an actual lawyer.


The email came on a Tuesday. A label wanted to license some of my AI-generated tracks. The operation looked legitimate. The conversation was professional. And the number they led with was 85% — 85% of royalties to me, 15% to them.

I read that number and felt something warm settle in my chest. That's a good split. I've read enough about the music industry to know that 85% is not a number labels throw around carelessly. My first instinct was to reply and work out the details.

I didn't. Something made me print the contract first.


I sat down with it that evening. Fifteen pages, standard formatting, lots of defined terms. The first few pages were fine — parties, definitions, the scope of what was being licensed. Professional language. Nothing alarming. I was reading the way you read something when you've already half-decided to say yes: looking for confirmation rather than problems.

That changed somewhere around page five.

I was looking for the definition of the royalty calculation base — the specific thing that 85% was calculated from. Gross revenue? Net revenue after deductions? It mattered. A label that deducts its marketing costs, distribution expenses, and promotional investments before calculating your share can make an 85% deal pay out at 40% in practice, or less. The headline number only means something if you know what it's a percentage of.

I found the royalty formula. I found the percentage. I did not find a definition of the base.

I went back and read it again. Same result. There was a term used throughout the royalty section — a term that the whole calculation depended on — that wasn't actually defined anywhere in the document. It was just used. As if it were obvious. As if we already agreed on what it meant.

We hadn't agreed on what it meant. We'd never discussed it.

I made a note. Moved on. Told myself maybe it was an oversight, maybe it was standard language, maybe I'd ask about it later. But something had shifted. I was reading differently now.


The sublicensing section came next.

When a label distributes your music, they have to pass usage rights to third parties — streaming aggregators, platform partners, distributors. That's how it works. You can't realistically expect them to come back to you with a signed letter every time they add your track to another service. Some degree of automatic sublicensing is built into how the whole system operates.

What matters is where that automatic zone ends and the "ask the author first" zone begins. Use your track in a film? Advertisement? Let someone remix it? Those feel like things that should require explicit, separate consent.

In this contract, the line wasn't clear. The category of uses that could be passed along without checking with me was wider than I expected. Not dramatically so — nothing jumped out as obviously wrong. It was more like a blurriness. A looseness in the language. The kind of looseness that, if it ever needed to be interpreted by someone, would probably be interpreted in the direction that was more convenient — and more convenient was unlikely to mean more convenient for me.

I kept reading.


And then I hit the clause about what happens when the contract ends. And I re-read it three times.

The deal had a multi-year term. When the term ends, you'd assume: relationship over, music back to me. But the clause said something else. The sublicensing agreements the label had made with third parties during the term — those continue. They don't end because my contract ended. They run for whatever term was specified in those third-party deals.

Third-party deals I was never a party to. That I'd never seen. That I had no visibility into whatsoever.

Think about what that actually means. I sign. Four years go by. The label, during those years, sublicenses my tracks to streaming distributors, platforms, whoever. When year four ends, I say I'm done. We shake hands — metaphorically — and go our separate ways.

Except my music doesn't go anywhere. It keeps circulating through agreements the label made, for however long those agreements last. I'm formally free. My catalog isn't.

I sat there and thought about that for a while. The image that kept coming to mind was a door. I thought I was being offered the option to walk through it eventually. What I was actually being offered was permission to stand next to it while someone else held the key.


I kept going. And I found the liability clause.

If any third-party claim arose in connection with the rights I'd transferred — a copyright dispute, a Content ID conflict, anything — I was responsible for resolving it. At my own expense. And if the label suffered losses during the time that dispute was being resolved, I was obligated to compensate those losses.

In full. No cap. No ceiling.

I put the contract down. I got up and walked away for a few minutes.

I want to be fair here: I don't think the label intended to use this clause as a weapon. Most deals don't end in disputes. Most of the time, a clause like this sits in a document forever and never matters. The people on the other end were perfectly pleasant throughout our entire conversation.

But a contract is not a conversation. A contract is what governs things when the conversation has broken down — when something has gone wrong, when people who were perfectly pleasant have stopped being pleasant, when there's money at stake and someone needs to assign the loss somewhere. And in that document, the place the loss was assigned was me.

Say a third party files a claim against one of my tracks. These things can take months to resolve. During those months, the label argues they couldn't monetize that track, and that they're losing income while the dispute sits unresolved. They bill me for it. All of it. No ceiling. And nothing in the document said that couldn't happen.

I'm not saying it would have happened. I'm saying the paper let it happen. And those are different things that matter in different ways.


After that, I looked for the exit.

What I wanted to find: a specific clause giving me — the author, the person who made the music — the right to walk away early if things weren't working. If the results were poor. If the relationship had soured. A defined mechanism. A real exit.

What I found instead was a provision saying that termination was available under applicable law. Which sounds fair until you think about what "applicable law" means when one party has professional legal representation and the other is an independent creator who has never litigated a commercial licensing dispute.

It's not that the law offers no protection. It's that "whatever the law allows" is a much more expensive and uncertain thing to rely on than a clause in the contract that simply says: the author may terminate with X days' notice under Y conditions. The former is a fight. The latter is a door.

There was no door.


I went back to the beginning of the contract and read it through one more time.

And this time I wasn't looking at individual clauses. I was looking at the shape of the whole thing.

On one side: flexibility. The ability to sublicense broadly, to pass rights to third parties, to keep those rights circulating after the main agreement ends. On the other side: exposure. Unlimited liability for disputes. A term with no clean exit. A royalty base that was never precisely defined.

All the flexibility was theirs. All the exposure was mine.

I don't think that was malicious. I think it was probably just how their template was written — language that had accumulated over years and contracts, that worked well enough for them, that they hadn't updated with the interests of the other party in mind. Maybe no one had ever pushed back on it. Maybe everyone who signed just looked at the 85% and stopped reading.

I pushed back by not signing.


If you're ever looking at a deal like this, here is what I'd look at before agreeing to anything:

What is the royalty percentage actually calculated from — and is that definition clearly written in the document, not just discussed in conversation?

What can the label pass to third parties without asking you, and what requires your explicit approval each time?

What happens to those third-party agreements if your contract ends early or at term? What specifically comes back to you, and when?

If a dispute arises — over rights, over Content ID, over anything — who pays to resolve it, and is there any limit on what you can owe?

Under what specific conditions can you, the author, end this agreement early — and is that written into the contract itself, not left to whatever the law might allow?

And then: show the whole thing to a lawyer. Not to confirm that it's fine. To find out what you're actually signing.

The percentage is the beautiful part. The part that makes you want to say yes before you've read anything else. It's supposed to be. But the percentage tells you almost nothing about what the deal actually costs you if something goes wrong — and a contract that gives one party every meaningful exit while leaving the other none doesn't become a good deal because the headline number is generous.

I read the contract. I said no. I still think that was the right call.


Related: How to Distribute AI Music to Streaming Platforms in 2026 · AI Music Generators in 2026 — 6 Compared · AI Music Watermarks Explained