AI music copyright still depends on human authorship

Using AI does not automatically strip a song or video of copyright protection. It also does not make every generated element yours to protect. The US Copyright Office's position turns on sufficient human authorship. Its January 2025 report announcement and May 2026 Senate testimony draw that distinction.

For an independent artist, the useful question is specific: what did you write, perform, arrange, or change in the finished work? A subscription receipt and a prompt history cannot answer that question alone.

Can AI-assisted music receive copyright protection?

Yes, human contributions can qualify when they meet the usual requirements. The Office says AI assistance and the inclusion of generated material do not automatically bar protection for a larger human-created work. The analysis concerns the human expression within it.

The Office identifies examples such as perceptible human-authored material, creative arrangement, and creative modifications. It does not say that every edit is enough. It also does not treat the mere provision of prompts as sufficient authorship of the resulting output.

Our analysis: this matters because a music release contains several creative decisions. Lyrics you wrote, a performance you recorded, and generated backing material do not have to share the same authorship history. A clear account can preserve the human contribution without claiming more than the evidence supports.

What the Copyright Office said in 2025 and 2026

The full copyrightability report applies existing copyright principles to AI outputs. It places weight on human control over expressive elements. It does not propose a new general copyright merely because a person paid for a generation.

The report also explains why prompting and authorship of an output are different questions. A prompt can express an idea or instruction while the system determines important expressive details in the result. Repeated attempts do not, by themselves, settle that distinction.

In May 2026 testimony, Register of Copyrights Shira Perlmutter restated the human-authorship requirement. She also reported that the Office had registered more than 7,000 claims containing disclosed AI-generated material. That figure concerns mixed claims; it is not a count of copyrights granted to wholly machine-authored works.

The testimony says the Supreme Court declined to review the Thaler case, leaving the appellate decision in place. Denying review is not the same as issuing a new Supreme Court merits opinion on every AI-assisted song.

Date	Primary document	What it establishes
March 16, 2023	Registration guidance.	How applicants should identify human and AI material.
January 29, 2025	Copyrightability report release.	The Office’s analysis of human authorship in AI outputs.
May 12, 2026	Senate testimony.	The Office’s stated position and reported registration experience.

Registration describes the work you actually made

The March 2023 registration guidance addresses works containing AI material. It instructs applicants to describe the human contribution and identify AI-generated content that is more than minimal. The guidance also addresses correcting a pending or completed registration when relevant information was omitted.

This is a disclosure and authorship issue. It is not an instruction to list an AI system as a human coauthor. Nor does excluding generated material from a claim necessarily exclude the human-created work surrounding it.

Our analysis: the best time to document the distinction is during production. Retrospectively reconstructing a project from its final stereo file can be difficult. A dated lyric draft and an original vocal session can answer questions that a finished export cannot.

Impact on users

Keep your writing drafts, performance recordings, project files, and meaningful revision history. These records help explain your actual contribution. They are not a promise that every element qualifies for protection, but they make an accurate assessment possible.

For a lyric video, preserve the approved lyric text separately from automated transcription. Note who wrote the lyrics and who made the visual sequence. A generator's involvement in one background image does not describe the authorship of the music, text, timing, and other visuals.

Consider this hypothetical production: a songwriter writes the lyrics, records a vocal, and uses generated instrumental material. An editor then builds a video with original typography and footage. The useful record identifies those contributions separately. It does not reduce the entire release to either “human” or “AI.”

When several people contribute, agree on accurate credits and relevant ownership terms. Human-authorship rules do not resolve a dispute between collaborators about who wrote a chorus. Contracts and the actual creative record still matter.

Commercial permission is different from copyright protection

Our analysis: a service can offer contractual permission to use an output without deciding whether copyright exists in every part of it. The ability to upload, monetize, or sell a track is therefore not a complete answer to whether you can stop another person copying it.

The reverse distinction matters too. A protectable human contribution does not clear somebody else's sample, lyrics, photograph, or voice. Your authorship and your permissions for outside material require separate attention.

This article concerns the US framework described by the linked sources. It does not determine ownership of a particular recording or promise the same result in another country. A release with a significant commercial dispute needs an assessment of the actual work.

A useful authorship record starts before export

Our practical recommendation is to keep a short production note with each final version. Describe the human source material, the tools used, the generated material retained, and the substantial changes. Make it specific enough that another collaborator can follow the history.

For example, “rewrote verse two and recorded all lead vocals” says more than “heavily edited with AI.” An accurate description serves the artist better than a large but vague ownership claim. It also helps ensure that a human contributor receives the credit they earned.

Our song-editing report explains why preserving originals matters. Our catalog metadata report covers another part of the same release record: keeping titles, versions, and credits attached to the right files.