The end of licensing as we know it?

The AI threat to artists, labels, and everyone in between that keeps me up at night

Previously, music rights holders could license music to streaming services for revenue and renegotiate terms. AI, however, is set to disrupt this model significantly.

Oct 9, 2024, 5:00 PM

Up until a few years ago if you owned the rights to music, you could license that music — say to a streaming service — in exchange for a share of the revenue generated. You would also have the power to renegotiate those licenses regularly. But AI is poised to radically change this dynamic, and, surprisingly, that’s not the most talked about aspect of AI’s potential impact on the music industry.

The Bait: Your Music = Data

Let’s assume that I’m an AI company called LarkGPT. I want to use your music to make new music that sounds like you. Not exactly like you, but enough that a fan might think it’s your latest song.

Let’s also assume that I am morally flexible.

I might scrape your music from the internet without paying for it, then turn your music into math (data) that my AI can alter just enough to create something new.

LarkGPT would then call this LarkGPT-1, which I could offer to the public in exchange for money, also letting others make music that sounds a lot like you.

So you come to me and say, “Hey, you need a license to use my music to make money!” Remember, morally flexible but not stupid, so I say, “Yeah, you’re right, let’s do a license! Here’s some money in exchange for your work. Let’s renegotiate this in three years, just like the streaming services do.” You say, “Great, thanks! I like money.”

The Switch: Your Music = My Synthetic Data

You have some money, and I have legally turned your music into math, aka my data. LarkGPT-1 is a legit company — even “fairly trained,” if you will. I have made millions from your work, so now it’s time to train my brand new AI LarkGPT-2. And trust me folks, it’ll be even better!

When it comes time to renegotiate our licensing deal, my lawyers email your lawyers and say, “Hey, we love your music and everything, but we don’t need this license with you any longer. When your music was turned into math, aka data, it was synthesized. Now our shiny new LarkGPT-2 model doesn’t need your music anymore; we only need our synthetic data.”

Wait, what?

“We appreciate working with you, and we’re willing to give you a little money because we think you’re great. But we have what we need to keep doing our thing. Thanks again for helping us build our global megacorp. And remember, we are a “fairly trained” company — we did pay you for your work.”

6 things you must do right now: PROTECT. CONTROL. MONETIZE.

PROTECT:

Beat the LarkGPTs to the punch by registering your works in SUREEL AI.

1. Turn your music into your own mathematical training data. Once your music (lyrics, artwork, voice, etc.) is in the Sureel AI Registry, you own and control the data versions of your work. Do this now!

2. Show the AI companies where the data versions of your work are. This is done via APIs and will allow them to identify your work. Here’s the important part: you must tell them that this is the ONLY place they are allowed to access your work for training. No other website, app, social media site, database, etc, is authorized!

This may sound backward, but by showing the LarkGPTs of the world your work, you are showing them what to avoid should they come across it as they scour the web, or even in their own databases. Remember, these companies don’t actually know what your work is.

CONTROL:

Now that your music is in a machine-readable format, and you’ve told LarkGPT that this is the ONLY place they are allowed to access your work for training, you can:

3. OPT-OUT by default or OPT-IN on your terms. You can now select which of your songs, lyrics, stems, and artwork, if any, you want to make available to LarkGPT, now and in the future. Once in the registry, you have a way to show them, in a language that computers can understand, what is OK for inclusion and what is not.

4. MAKE THE RULES. In a language that LarkGPT understands, you can now set the rules for usage in their models. For example, you might say, “I don’t want my voice to be altered by more than 50% because I want what you make to always sound like me!” or “Don’t create anything that uses more than 50% of my music, so my work retains its individuality.” You can also let them know when you are not OK with your music being used. No political ads on the internet, anyone?

Now that the LarkGPTs of the world know what your works are — as they are already in training format — they are actually incentivized to do the right thing, because they will A) know how to, and B) be getting less expensive and higher-quality data for their future models, therefore minimizing future liability on their end. (Oh, and you can check via your real-time dashboard to make sure they are doing the right thing. Trust but verify, right?)

MONETIZE:

The traditional licensing model makes little sense if you can’t put a value on your work because you don’t know what’s being done with it.

But if you’ve gone through steps 1–4, you now have a real-time dashboard that shows you how your works are being used, and also how much your work influences what LarkGPT creates.

5. License your music as training data and share the revenue the AI makes. Each time LarkGPT creates something with your work, it can credit you and pay you according to the rules you set in step 4 above.

6. License based on how much your work influences the whole AI model. You can now see in real time when and where your music is used, and what it is being used to create.

This is where it gets big because these companies are already worth billions of dollars more than the entire music industry. And there are over 800 of them.

PROTECT even if you don’t want to CONTROL and MONETIZE

At the heart of the matter is a basic proposition. If you don’t want your music in AI models like LarkGPT, then you should have the right to opt out of them. But if you don’t have a way to show them what your music/models are, then they won’t know how to do as you wish.

AI is fundamentally different from any technology that’s come before. And though you might hate the idea of it, you cannot flat-license or sue your way out of it. These outdated business models play to LarkGPT’s advantage. Every day that goes by without registering your songs is another day that a new version of the model will be made.

LarkGPT-1 can only exist with your music, but LarkGPT-2, 3, and 4 need it less with each model — -and will eventually not need it at all.

If you’re an individual artist you can make money if you want, or leave it on the table if you don’t. That’s your call.

Suppose you’re a music company with a large catalog. Or perhaps you just bought billions of dollars worth of music assets. In that case, it’s a simple question: are the assets I own worth more registered and protected, or worth less?

So being morally flexible, I’d just LOVE to do a flat license with you for my next LarkGPT model. It’s much easier to pay you when you’re blind to what I’m doing with your music. I’ll take that model any day.

Music is now data, there’s no going back. The question is, who owns it?

Hey, all you film, TV, news, image, book, medical, fintech companies… you got some good looking data… how about we do a nice flat license?

Full disclosure: I have joined Sureel AI as Co-President, along with Aileen Crowley, and have had a chance to see this incredible technology in action. It’s important to note that Sureel AI’s registry is non-exclusive, and free to enter and opt out of training (up to 100 assets). This means that if other registries exist (like the recently-announced SoundExchange registry), there is no harm in adding your work to them all. It’s imperative you get your work registered immediately, as it “starts the clock.” Sureel operates in a way unlike any other company we’ve seen. (Sureel CEO Tamay Aykut will be writing the technical version of the above, which I will then share.)

There is nothing to lose by opting in, and much to lose by waiting. What you create is too valuable, and this technology moves too fast.

Benji Rogers

Co-President

Published: