# Can you copyright AI-generated art? What the US Copyright Office says

Source: https://creatisimo.net/topic/can-you-copyright-ai-generated-art-what-the-us-copyright-office-says/
Site: Creatisimo, category AI image tools and generators
Published: 2026-10-29. Updated: 2026-10-29. Replies: 0.

## Question by Mira Talbot (staff), 2026-10-29

Not purely AI-generated output. The US Copyright Office says copyright needs human authorship, and that prompts alone do not give enough control. You can claim the human parts: your own artwork visible in the output, your creative selection and arrangement, and your creative edits. This is general information, not legal advice, and it covers the United States only.

## The Office's position in one place

The Office published Part 2 of its AI report, on copyrightability, on January 29, 2025. The report, [Copyright and Artificial Intelligence: Copyrightability](https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf), sets out its conclusions as a short list. The ones that matter to image makers:

- Questions about AI and copyrightability can be resolved under existing law, without new legislation.
- Using AI tools to assist, rather than stand in for, human creativity does not affect whether the output can be protected.
- Copyright protects original expression by a human author even if the work also includes AI-generated material.
- Copyright does not extend to purely AI-generated material, or to material where human control over the expressive elements is insufficient.
- Whether a human contribution is enough is decided case by case.
- Based on how current generally available technology works, prompts do not alone provide sufficient control.

On that last point the report reasons that prompts act as instructions conveying ideas, and that the system, not the user, is largely responsible for determining the expressive elements in the output. It adds that this could change if technology gives users more control, and the Office says it will keep monitoring developments.

## What counts as human authorship

The report separates three kinds of human contribution beyond the prompt.

| Contribution | What the report says |
| --- | --- |
| Your own work used as input | If your copyrightable drawing or photo is perceptible in the output, you are the author of that portion, with protection limited to your perceptible expression |
| Selection and arrangement | A creative selection, coordination or arrangement of AI-generated material can be protected as a whole, though not the AI-generated elements standing alone |
| Modifying the output | Creative edits that add new authorship are protected, assessed case by case; the Office notes tools such as region editing give more control over placement of elements than prompts do |

The report gives examples from registrations. In the case of the graphic novel Zarya of the Dawn, registration covered the human-authored text and the selection, coordination and arrangement of text and images, but not the Midjourney images themselves. For an image called Rose Enigma, an artist started from their own hand-drawn input and the Office registered the work with a note limiting it to the unaltered human pictorial authorship clearly perceptible in the deposit. In contrast, the Review Board upheld a refusal for an image made by applying a painting style to a photograph with an AI app, finding the applicant exerted insufficient creative control.

The report also notes a basic point: including AI-generated elements in a larger human-authored work does not affect the copyrightability of the whole. Its example is a film that uses AI-generated special effects.

## Practical steps if you want protection

1. **Keep your process files.** Sketches, layered files, version history and notes are the evidence of what was human.
2. **Make real edits.** Painting over, compositing and recomposing show authorship that a single prompt result does not. A selection of only two or three elements is, per a footnote in the report, generally not enough.
3. **Disclose AI content when you register.** The Rose Enigma example shows applicants disclaiming non-human expression, and the registration was limited accordingly.
4. **Register the human parts you can name.** Be specific about what you made versus what the tool made.
5. **Do not treat tool terms as a substitute for authorship.** A vendor's terms govern what you may do with its output. The Office's test is about human authorship, and a contract between you and a vendor is not part of that test. For the licensing side, see the [Midjourney or Firefly discussion on client work](/topic/midjourney-or-firefly-for-client-work-which-one-can-you-actually/).

## What this does not settle

- **Licence versus copyright.** Whether you may use an image commercially under a vendor's terms is a different question from whether you own copyright in it.
- **Other countries.** The report's international section says countries that have addressed the issue so far agree that copyright requires human authorship, but rules differ in detail, so check where you sell.
- **Court cases.** The report cites a district court decision, in Thaler v. Perlmutter, holding that human authorship is a bedrock requirement of copyright. The Office's [AI landing page](https://www.copyright.gov/ai/) now also lists a Court of Appeals decision affirming the refusal of registration in that case. Check that page for later updates before relying on this thread.
- **Training data.** Part 2 deals with copyrightability of outputs, not with whether training on existing works is lawful.

For a project that mixes methods, the [book cover process thread](/topic/book-cover-process-hand-sketch-ai-color-studies-final-paint/) is a good example of how a hand sketch and a final paint leave a clear human trail.

---
Cite as: Creatisimo, "Can you copyright AI-generated art? What the US Copyright Office says", https://creatisimo.net/topic/can-you-copyright-ai-generated-art-what-the-us-copyright-office-says/, accessed 2026-10-29.