Your Music Is Already Training AI

Estimated reading time: 7 minutes

There’s an artist I love named Ximena Castro. The artist has a great voice and amazing protest lyrics about injustice in Iran, Gaza, ICE, and immigration. It’s very current, emotionally powerful stuff. One day when I was scrolling, it got me thinking: How is she cranking this music out so fast? The punchline? Ximena’s music is AI, or at least heavily AI-assisted. And honestly, it’s incredible.

AI music isn’t coming anymore. It’s already here. Millions of songs a month are being AI-generated. Much of it is mediocre, sure, but in the hands of the right artist or the right prompt engineer, it could be damn good already. Now, just saying that may get me some hate mail from anti-AI purists whose main complaint is that AI has no soul. And look, I understand that concern, but don’t forget that there’s plenty of mediocre music being made by humans, too.

Table of Contents:
• Music and AI learning
• Why artists should educate themselves
• Copyright protects expression, not style
• Should artists opt out of AI?
• What should artists do?

Music and AI learning

The truth is that your music is already being used for AI learning. But what do distributors and streaming platforms allow when it comes to AI scraping and AI learning? Where is all this headed with licensing and monetization, and most importantly, what should you actually do about it as an artist and songwriter?

Whether artists realize it or not, AI learning may already quietly have become one of the biggest monetization fights in music. Before I go any further, let me double down on something I already hinted at. Love it or hate it, AI music is here to stay. You can try to fight it completely, but I think you’re going to lose that battle. Recently, Suno’s CEO compared AI music to Ozempic. Nobody likes talking about it, but everyone’s doing it right now.

Why artists should educate themselves

The most important thing you can do with regard to AI as an independent artist is to educate yourself: AI rights, AI learning, AI monetization, and how to protect yourself as a creator. So, let’s start with the role that your distributor plays with regard to AI music. When artists ask whether distributors like DistroKid, TuneCore, and CD Baby allow AI-generated music onto Spotify and Apple Music, the answer is mostly yes. Across the board, the big distributors are basically saying, “If you created it, or if you have the rights to it, you can distribute it.” So, if you use AI tools to help generate music, artwork, vocals, whatever, generally speaking, that’s fine. What’s not fine is impersonating another artist, cloning someone’s voice without permission, or infringing on someone else’s copyrights.

That part is actually fairly straightforward. AI music is largely being treated under the same copyright framework as traditional music, and it’s already here in a big way. What I think artists should be paying attention to is what happens after your music enters the distribution ecosystem. Because when you upload music through distributors, you grant broad licenses through their terms of service. The exact wording varies, but generally speaking, distributors reserve the right to sublicense your content to platforms, support different downstream music uses, enable user-generated content, and sometimes allow remixing or derivative formats to be created. Now, maybe in the future that means nothing. Or maybe someday your distributor signs a deal with an AI remixing company or some AI music platform, and suddenly those broad rights that you granted become much more significant.

One thing distributors generally do not say in their agreements is: “We will protect your music from AI learning.” The reason is actually fairly simple. Distributors are pipes. Their job is to move your music from their service servers to Spotify, Apple Music, TikTok, YouTube, and everywhere else. They are distribution infrastructure. They’re not AI gatekeepers policing how the internet learns from content. And frankly, they distribute music, but they don’t put it online for public consumption themselves. So, there’s nothing much for AI bots to really scrape at the distribution level. So, what about the big streaming platforms? Spotify, Apple Music, YouTube. These platforms mostly regulate what gets uploaded. They care about copyright infringement, fraud, impersonation, deepfakes pretending to be Drake or Taylor Swift. But that’s very different from controlling what AI systems learn from.

Maximize your audience and earnings with our distribution services

Copyright protects expression, not style

This is the key distinction artists need to understand: copyright protects expression, but it does not protect style. An AI system is not allowed to literally copy your song, but it can absolutely learn from your chord progressions, your production style, your vocal textures, your genre conventions, and that’s the loophole the entire AI industry is currently built around. And nowhere is this more obvious than on YouTube. YouTube is basically one giant AI training ground. Officially, YouTube prohibits scraping and large-scale downloading in its terms of service. But in practice, much of YouTube has already been scraped over the years, helping build the enormous datasets that these AI systems learn from. And YouTube’s Content ID system, which is incredibly sophisticated, was designed to identify copies of recordings. It can tell if someone uploaded your exact song or even a tiny snippet of it and used it in their video.

But it cannot tell if an AI model learned your style. And it definitely does not compensate you because your music helped train a model. At least it doesn’t today. Now, as I’m sure you know, the AI companies are currently getting sued by major labels and publishers over copyright infringement, and the law is still catching up here. But let me be clear about something: The music industry is not trying to stop AI. It’s trying to control it, monetize it, and integrate it. That’s where all of this is headed. Right now, AI companies have mostly operated under a “train first, deal with legal later” philosophy. But over time, I think you’re going to see licensed training datasets, verified catalogs, opt-in systems for rights holders, attribution models, and eventually some form of revenue sharing tied to AI usage. Think of it almost as content ID evolving into something broader and more sophisticated.

Should artists opt out of AI?

It’s not just, “Did someone copy your song?” but, “Did your content help train this system?” We’re early still, and nobody really knows yet how compensation for AI learning is ultimately going to work. Which brings us to the big question artists should keep asking: Should I opt out of AI? And honestly, I just don’t think most artists will realistically be able to even opt out. Look at YouTube. Even if you don’t upload your music, music spreads. It gets shared. It gets reposted. It gets analyzed. It influences systems indirectly. So, I actually think the more useful question is, “How do I position myself intelligently for the world that’s coming?” Because the real choice may not end up being opt in versus opt out. It may become participate and potentially get compensated, or sit outside the system entirely, if that’s even possible in today’s connected world.

I understand why that makes artists uncomfortable. But just like the internet and sampling and drum machines and home recording before it, AI is not going away. It’s going to become deeply integrated into how music gets made.

What should artists do?

First things first: Own your rights whenever possible. Your master rights, your publishing rights… because if future AI licensing deals happen, and I’m certain they will, the money is going to flow to whoever owns the rights. And if you don’t own them, you don’t participate.

Second, start treating your catalog like an asset, not just like content, an actual asset with financial value. Copyright your music. Archive your sessions. Keep your stems. Organize your files. Back them up. Make sure your metadata is accurate. Metadata matters more than most artists realize because if future systems can’t properly identify you, you don’t get paid.

Third, build direct fan relationships. Ironically, AI will make human connection more valuable — not less. AI is going to flood the market with more music and more noise, which means fans will increasingly value the things AI can’t truly replicate: connection, trust, authenticity, story, identity.
And finally, use AI as a tool. Don’t hide from it. The artists who win over the next decade are probably going to use AI to move faster, to market themselves better, to test more creative ideas, to get unstuck creatively, and to create more efficiently.

That’s my biggest piece of advice when it comes to AI. Use it as a tool, but don’t outsource your identity to AI. I’ve seen too many people write and create generic content using AI and lose their own voice, lose their soul. Because if you want to stand out as an artist, the thing that will matter most is still what has always mattered the most: you, your perspective, your humanity, your story, your taste. AI can imitate style, but it can’t be you… unless you let it.

You can watch the original version of this blog as part of our Indie Music Minute series on the Disc Makers YouTube Channel.