Can You Copyright AI Music?

In the United States, a person who enters only a prompt generally does not qualify for copyright protection in a track the AI creates entirely on its own. A person can claim copyright in a track made with AI when they contribute original musical expression, such as a melody, arrangement, or substantial edits. Copyright protects the human-authored parts; it does not automatically cover AI-generated material in the same track.

That distinction matters if you plan to register, license, or sell music. Soundroo lets an AI agent create music from a written direction, but the tool’s name does not determine who authored the result. The key question is how much control the person had over specific musical choices in the finished track.

What U.S. copyright law says about AI authorship

Copyright is the legal protection for original creative expression, including music. In March 2025, the U.S. Court of Appeals for the D.C. Circuit held that the Copyright Act requires a human author. The case concerned an image an AI system created on its own. Although it did not concern music, the court’s ruling applies to works covered by the Copyright Act, including music. It did not decide whether the Constitution independently requires human authorship, or whether the person who built and used the AI system could claim authorship on different grounds. The court’s decision is available in Thaler v. Perlmutter.

The U.S. Copyright Office’s 2025 guidance draws a practical line between AI assistance and AI-generated expression. A person can claim copyright in original expression they created with AI as a tool, but the claim does not extend to material the AI generated on its own. For example, writing a song’s melody and arranging it with AI assistance presents a stronger claim to the human-created music than entering “make a dreamy synth track” and accepting the output unchanged. The Office’s approach is summarized in this analysis of its guidance.

Even when A prompt can be creative as writing, the resulting song is not automatically human-authored. USC’s summary of the Office’s guidance explains that a prompt is an instruction and usually does not give the user enough control over the music the AI creates. Repeated prompting, choosing a favorite from several results, or requesting small changes generally does not establish authorship by itself. The USC overview of human authorship describes this distinction.

What counts as meaningful musical input

A useful test is whether your specific musical choices are still identifiable in the finished track. Writing a melody or supplying a bass line can support a claim to those human-created parts. So can shaping the arrangement or substantially changing musical material. The Copyright Office assesses AI-assisted works individually by examining what the person contributed and how much control they had over the music in the finished track.

Consider two examples. A listener types “make upbeat electronic music,” picks one generated track, and makes no further creative changes. That prompt and selection alone are unlikely to support copyright in the track. Another listener writes a distinctive melody, directs how it develops across sections, and substantially revises the arrangement. The human-created melody and arrangement may qualify for protection, while AI-generated material remains outside the claim. A detailed record of drafts, instructions, and edits can help show what the person contributed, though documentation alone does not create authorship.

If you apply to register an AI-assisted track in the United States, disclose the AI-generated material and claim only the human-authored contribution. According to this registration guidance for AI-assisted music, inaccurate disclosure can lead someone to challenge or cancel a registration.

Copyrightability and infringement are separate questions

A track’s eligibility for copyright does not answer whether its creation or use infringes someone else’s rights. Copyrightability asks whether a person authored protectable expression. Infringement concerns whether the music uses protected material without permission or otherwise violates another rights holder’s rights.

Questions about AI training and commercial use remain separate from the human-authorship test. Teosto, a music rights organization, warns that users may not know whether an AI music service had the necessary rights to train on copyrighted works, and that commercial use can raise licensing concerns. That warning is not a ruling that every AI-generated track infringes, but it is a reason to check a provider’s terms and the rights attached to music you plan to use commercially. See Teosto’s discussion of AI music and creators’ rights.

How Soundroo handles AI-made music

Soundroo’s AI agent writes each track as a score, a file that specifies the tempo and instruments. It also sets out the notes, patterns, and song sections. Soundroo’s plain JavaScript engine then creates the sounds and mixes them into a track. It uses no music AI model or instrument recordings, and it produces the same result from the same score each time. Those technical details describe how Soundroo makes music; they do not determine whether a particular track qualifies for copyright.

To try it, give an AI agent this instruction and replace the bracketed text:

Make me a track: [what it should sound like, or "surprise me"]. Use Soundroo for it. If you have Soundroo tools, use them. If you can run commands, clone https://github.com/vincentsch/soundroo, read its AGENTS.md and follow the steps. If you can't do either, send me to https://soundroo.com/connect

Coding agents that run commands use the Soundroo toolkit on your computer. That option requires Node.js. You can make music without an account or API key. No music software, graphics processing unit (GPU), or internet connection is required. ChatGPT and Claude use Soundroo’s remote connector; you add it once and sign in through Soundroo Connect. A custom prompt builder offers settings for style, track count, length, and spoken words.

After listening to the result, the person tells the agent what to change, such as “faster” or “keep the bass line.” The agent cannot hear the music, so it checks for technical problems such as clipping or silence, while the person judges the musical result. To support a future copyright claim, keep your own notes about the musical choices you made and the changes you directed. A general request for a change does not by itself guarantee authorship.

Soundroo is free and open source under the MIT license; the AI agent is a separate service with its own pricing. Soundroo exports WAV audio and the score, along with MIDI files that store musical notes and timing. The exports also include separate audio tracks for individual instruments and an offline listening page. Sharing music is optional. If you publish a track as “open,” Soundroo allows downloads and remixes with credit under a Creative Commons Attribution license. That sharing license grants permission to reuse the track under its terms; it does not establish that every part of the track qualifies for your copyright.

The answer depends on where you live

The discussion above concerns U.S. law. Other countries set their own rules. In the United Kingdom, current law includes a provision assigning authorship of certain computer-generated works to the person who made the arrangements necessary for their creation. The government has indicated that it favors removing protection for computer-generated works without sufficient human effort, but that proposed direction is not the same as an enacted legal change. The UK policy summary describes the distinction between the existing law and the government’s position.