Who owns AI-generated music?
This question gets answered confidently by a lot of people who are selling you something. The confident answers are usually wrong, and the wrong version can cost you a release.
Here is the position as it actually stands, with the parts that are settled separated from the parts that are not.
Copyright in the US requires a human author
The US Copyright Office has been consistent: a work must have human authorship to be registered. Material generated purely by a machine from a prompt does not qualify, and the Office has refused registrations on exactly that basis.
This is not a statement about whether AI music is legitimate. It is a statement about what the copyright register will accept. A purely prompt-generated track may have no copyright at all — not yours, not the platform's, not anyone's.
A commercial licence is not ownership
Most generation platforms grant paying users a commercial licence to use the output. That is a contract between you and the platform. It governs your relationship with them; it does not create rights against the rest of the world.
Suno's own terms, for instance, state that it makes no representation or warranty that any copyright will vest in the output. Read that carefully: the platform selling you AI music is explicitly declining to promise you own it.
A licence says "we will not sue you for using this". Ownership says "you can stop other people from using this". They are not the same thing, and only one of them is what you need to exclusively control a release.
Where the line actually falls
The more of the work is yours, the more solid your position is. This is a spectrum, not a switch.
| What you did | Where you stand |
|---|---|
| Recorded and performed it; used AI to mix or master | Yours. Processing does not affect authorship. |
| Wrote and sang it; AI generated the backing | Your vocal and composition are yours. The generated parts are the unsettled bit. |
| Prompted a full track, then edited it substantially | Your creative choices may be protectable. The underlying generation is not. |
| Typed a prompt, released the output as-is | Weakest position. Likely no registrable copyright. |
What this means practically
If you are releasing commercially, the safest material is music where you made the substantive creative decisions and used AI as a tool — the way you would use a compressor or a session player.
Distributors increasingly require AI disclosure, and streaming services are building detection. Undisclosed fully-generated material is a growing risk to your account, separate from the copyright question.
If a track matters to you commercially, the part you should be most careful about is not the mixing. It is whether you can point to the human authorship.
What dhun does about it
We take no rights in your music — no cut, no publishing share, no licence to your masters. Running a compressor on your song does not entitle us to it.
We also do not tell you that you own 100% of everything you make here, because for generated audio that would be a promise we cannot honour. Our terms draw the line where it actually falls rather than where it would sound best.
For the audio you bring us — your recordings, processed by our tools — the result is unambiguously yours.
This is general information about how copyright currently treats AI-generated material, not legal advice. If a specific release matters commercially, talk to a lawyer who practises in this area.