South Korea Now Lets You Copyright AI-Assisted Music — If You Did the Real Work
KOMCA reversed its ban on AI-assisted music registration, but requires strict disclosure. Here's what creators need to know about the new rules.
South Korea’s largest music copyright collective just flipped the script. The Korea Music Copyright Association (KOMCA) announced it will now register AI-assisted musical works, reversing a previous ban that had blocked such registrations entirely. But there’s a catch that changes everything: you have to prove a human did the heavy lifting.
This isn’t a free pass for AI-generated slop. KOMCA’s new standards require that a human played a “substantial, leading creative role” in making the music, and the registration application has to tell a detailed, accurate story about how the work was actually created. False or sloppy filings carry real consequences — royalty holds, clawbacks, and possible termination of the trust agreement.
Why This Matters Beyond Korea
KOMCA’s shift is significant because it’s one of the first major copyright collectives anywhere in the world to create a formal pathway for AI-assisted works. Most jurisdictions are still arguing about whether AI-generated content can be copyrighted at all. South Korea just skipped ahead to the harder question: how do you register it?
The practical implications are substantial. KOMCA manages royalty distribution for a huge portion of South Korea’s music industry. When an AI-assisted song gets registered, it enters the same royalty pool as traditionally composed music. That means the people who created it — assuming they meet the human-creative-leadership requirement — can earn royalties from streaming, radio play, and licensing just like any other registered work. For independent musicians who use AI tools as part of their production process, this could be the difference between earning nothing and earning something.
The policy doesn’t actually change South Korean copyright law. KOMCA doesn’t have that authority, and lawmaker Kim Jae-won publicly criticized the move, pointing out that the National Assembly hasn’t amended the copyright statute to give registered AI-assisted works the same legal protection as traditional compositions. So right now, you can register with KOMCA, but that registration may not carry the full weight of copyright enforcement in court.
This creates an interesting middle ground. KOMCA handles royalty distribution for its members, so registration affects how money flows. But the legal copyright status of these works remains murky until the legislature acts. It’s a bit like having a driver’s license that’s valid for highway tolls but may not hold up in court if you get into an accident.
What makes this particularly relevant for the broader AI industry is the documentation requirement. KOMCA isn’t just asking “did a human make this?” — they’re asking “prove it, in detail, with records.” This sets a precedent that other copyright offices around the world may follow. If you want to claim copyright over AI-assisted work, you’ll need to show your work, literally.
The Global Context: Everyone Is Figuring This Out
South Korea isn’t operating in a vacuum. The same week KOMCA made its announcement, several other developments reshaped the AI music landscape.
In the United States, the legal battle between music labels and AI companies Suno and Udio continues. Both companies argue that using copyrighted music to train their models falls under fair use. The outcome of these cases could set precedent for how AI training data is treated across all creative industries, not just music. If the courts decide that training on copyrighted data is fair use, it fundamentally changes the economics of AI development. If they decide it isn’t, every AI company that trained on copyrighted material faces potential liability.
The Music Artists Coalition, a nonprofit advocacy group, has been pushing a framework built on three principles: informed creator consent, fair compensation, and transparency in how AI uses are authorized and monetized. Executive director Ron Gubitz put it plainly: “We’re not anti-AI. We think that the clarity, consent, and compensation framework does provide us the opportunity to have the conversations and make those agreements that work.”
This framework matters because it shifts the conversation from “is AI theft?” to “how do we make AI work for creators?” That’s a more productive framing, and it’s one that KOMCA’s policy change implicitly supports. By creating a pathway for AI-assisted works to earn royalties, KOMCA is essentially saying: AI is here, let’s figure out how to make it pay.
Meanwhile, the European Union’s AI Act, which took effect August 2, requires AI system providers to watermark AI-generated audio, images, text, and video outputs. That’s a transparency mandate, not a copyright solution, but it’s part of the same global push to figure out how AI-generated content fits into existing legal frameworks. The watermark requirement specifically targets AI-generated content, which raises its own questions about where the line is between “AI-generated” and “AI-assisted.”
The Warner Music union also ended its litigation over AI recordings this week, suggesting that at least some parties are finding resolution through negotiation rather than courts. Whether that’s a good sign or a sign of capitulation depends on who you ask.
What Creators Should Actually Do
If you’re a musician using AI tools in your workflow, KOMCA’s new rules give you a template that’s worth following even if you’re not registering in South Korea:
Document everything. Keep project files, revision history, and notes about which parts were AI-generated versus human-created. The more detailed your documentation, the stronger your position if your work is ever challenged. This isn’t just about copyright registration — it’s about being able to prove your creative contribution if someone questions it.
Be honest about AI’s role. KOMCA’s requirements for disclosure are strict, and the penalties for false filings are real. This isn’t the time to minimize AI’s contribution to save face. Transparency protects you. If you used AI to generate a melody but wrote all the lyrics and arranged the production yourself, say so clearly. The documentation requirement isn’t a hurdle — it’s a protection.
Understand your local rules. KOMCA’s policy is specific to South Korea and its royalty distribution system. Your jurisdiction may have different requirements for copyright registration, and the legal status of AI-assisted works varies widely from country to country. In the US, the Copyright Office has generally refused to register works created entirely by AI, but has allowed registration for works where AI was used as a tool under human direction. The details matter.
Follow the money. The real fight isn’t about registration — it’s about who gets paid when AI uses your work. The consent and compensation frameworks being developed by organizations like the Music Artists Coalition are more directly relevant to most creators than any single country’s copyright registration policy. If you’re a session musician whose recordings were used to train an AI model without your permission, the KOMCA registration rules don’t help you. The lawsuits and advocacy frameworks do.
The Bigger Question
KOMCA’s move raises a question that the industry hasn’t answered yet: if a human used AI as a tool — the way they might use a synthesizer, a sample library, or a DAW plugin — does that make the output “AI-generated” or “human-created with AI assistance”?
The answer matters enormously. If AI assistance automatically disqualifies a work from copyright protection, you’d also have to strip copyright from music made with Auto-Tune, algorithmic mixing tools, and AI-powered mastering services. The line between “AI tool” and “AI creator” is blurrier than most policy discussions acknowledge.
Consider a concrete example. A producer uses an AI tool to generate ten melody options, picks one, writes lyrics over it, records vocals, arranges the production, mixes and masters the track. Is that AI-generated? The melody came from AI, but everything else was human-directed. KOMCA would likely consider this registrable, because the human made the creative decisions. But what if the producer just picked the first melody option and released it with minimal changes? That’s a harder call, and the documentation requirement is supposed to help draw that line.
South Korea’s approach — requiring human creative leadership while allowing AI as a tool — is one attempt to draw that line. It’s imperfect, and the legal infrastructure isn’t fully in place yet. But it’s a starting point, and at this stage, an imperfect starting point is better than no framework at all.
The other question is whether this approach scales. KOMCA is a national collective with specific jurisdiction and membership. The global music industry doesn’t have a single copyright authority, and the fragmented nature of international copyright law means that a work registered in South Korea may not enjoy the same protections in the US, EU, or elsewhere. For now, creators will have to navigate a patchwork of national rules, each with its own definition of what counts as “human-created” versus “AI-generated.”
The fight over AI and copyright is just getting started. What KOMCA did this week won’t be the last word, but it’s an early signal of where things are heading: toward a world where AI is part of the creative process, and the rules have to catch up. For now, the best thing creators can do is document their work carefully, understand their local regulations, and stay informed as these frameworks continue to evolve.
Reader Dispatches
No reader responses yet. Share your technical perspective below.
Related Posts
Suno Lost to GEMA. Now Every AI Company With EU Users Has a Problem.
A Munich court ruled Suno infringed GEMA copyrights and asserted jurisdiction over US-based AI training. The implications reach far beyond music.
AI Image Generators Compared: Midjourney vs OpenAI Image Generation vs Stable Diffusion
How Midjourney, OpenAI's image generation in ChatGPT and the API, and open-weight Stable Diffusion differ in workflow, control, licensing and cost model, based on official documentation, with a guide to choosing.
89 of Douyin's Top Dramas Are Now AI-Made
DataEye says 89 of Douyin's top 100 dramas in May were AI-generated. China's Q1 short-drama flood is 95% synthetic. The hit rate is brutal.