Are AI images copyrighted? A 2026 guide for designers and clients
What the US Copyright Office, the courts, the EU AI Act and tool terms say about AI images in 2026, with a practical checklist for designers and clients.
Published 10 min read
Are AI images copyrighted? In the United States, an image generated from a text prompt alone is not protected by copyright, because the law protects only human authorship. What a person adds can be protected: their own drawing, their edits, the way they select and arrange material. Other countries are still working the question out, and at least one court has taken a different view.
That is only half of what designers and clients need to know. Who may use an image is set by the tool's terms and your plan, not by copyright. Whether the image copies someone else's work is a third question. And for logos, trademark law matters more than any of them.
This guide covers what the US Copyright Office and the courts have said, how other places differ, what I could and could not verify for Saudi Arabia and Egypt, the EU's labelling rules, what the main tools' terms say, trademarks for logos, and a checklist for designers and clients.
This article is general information, not legal advice. For a real decision, speak to a lawyer in your country.
Three questions people mix up
| Question | What decides it | The practical version |
|---|---|---|
| Is the image protected? (copyright) | The law where you are, and how much of the work a person made | Can a competitor copy your generated poster? |
| May I use it? (permission) | The tool's terms, your plan and your client contract | Can you print it on packaging you sell? |
| Does it copy someone else? (infringement) | What the image shows, compared with existing work | Does it show a famous character or a real person? |
Most confusion online comes from mixing these up. "I own it" (from the terms) does not mean "it is protected" (by the law), and neither of them means "it is safe" (from someone else's claim).
The United States: the clearest official position
The Copyright Office report
On 29 January 2025, the US Copyright Office published Part 2 of its report on copyright and artificial intelligence, on copyrightability. Its main conclusions:
- AI output is protected only where a human author has determined sufficient expressive elements.
- Prompts alone are not enough.
- Protection can cover a person's own work that is still visible in the output, and a person's creative selection, arrangement or modification of the output.
- Using AI to assist, or including AI material inside a larger human work, does not block protection for the human parts.
Two registration decisions
- Zarya of the Dawn (21 February 2023). A comic book illustrated with Midjourney images. The Office protected the author's text and her selection and arrangement of text and images, but not the Midjourney images themselves.
- Théâtre D'opéra Spatial (September 2023). The prize-winning image was made with Midjourney, through many prompts and some Photoshop edits. The Office's Review Board refused to register it because the AI-generated material was more than minimal and was not excluded from the claim. The artist, Jason Allen, took the refusal to a federal court in Colorado in 2024, so check where that case stands before you rely on it.
The courts: Thaler v. Perlmutter
Stephen Thaler tried to register an image and named his AI system as its only author. In March 2025, a federal appeals court in Washington, DC agreed with the Office that copyright requires a human author. On 2 March 2026, the Supreme Court declined to hear the case, so that ruling stands.
If you register work in the US
The Office's registration guidance asks applicants to disclose AI-generated material that is more than minimal and to describe the human contribution. Keeping your process files makes that easy.
Elsewhere: no single rule
In November 2023, the Beijing Internet Court went the other way in one case, Li v. Liu. It treated an image made with Stable Diffusion as protected, pointing to the user's many choices of prompts and settings. One case is not a national rule, but it shows that the answer depends on where you are.
Saudi Arabia
Saudi Arabia issued a new Copyright Law in 2026 (Royal Decree M/169) to replace its old one. The law and its implementing regulations, published in the official gazette Umm Al-Qura on 31 July 2026, define the author as "the person who creates the work". They also allow copying lawfully published works to develop AI products and algorithms, under conditions such as keeping records of what was used. In those texts I found no rule on who owns an image that a tool generates.
Law firms have also reported 2025 guidance from the Saudi Authority for Intellectual Property (SAIP) on AI-related works, but I could not open an official copy of it. If protection matters to your project, ask SAIP directly or a Saudi IP lawyer.
Egypt
Egypt's main law is the Intellectual Property Rights Protection Law (Law 82 of 2002), and the Egyptian Intellectual Property Authority registers and oversees these rights. I could not find an official Egyptian guideline on AI-generated images. Until there is one, ask the authority or an Egyptian IP lawyer before you rely on protection for generated work.
The EU AI Act: labelling, not ownership
Article 50 of the EU AI Act applies from 2 August 2026. In general terms:
- Tool makers whose systems generate images, audio, video or text must mark the output in a machine-readable way, so it can be detected as generated. The European Commission's FAQ gives tools already on the market before that date until 2 December 2026 for this part.
- Anyone publishing a deepfake, meaning generated or altered images, audio or video that look like real people, places or events and could pass as real, must disclose that it is artificial.
- Clearly artistic, creative, satirical or fictional work can disclose this in a way that does not spoil the work.
These rules are about transparency, not copyright. They matter to you if your work reaches people in the EU, and they show where labelling is heading everywhere.
What the tools' terms say (checked October 2026)
Terms change, so read the current version for the plan you are on. Here is what the official pages said when I checked:
| Tool | What its official pages say |
|---|---|
| OpenAI (ChatGPT images) | As between you and OpenAI, you own the output. Output may not be unique, and other users may receive similar results. |
| Midjourney | You own what you create, to the extent the law allows. A company with more than US$1 million a year in revenue needs the Pro or Mega plan to own its images. Images are visible to others by default; Stealth mode is only on Pro and Mega. |
| Leonardo.Ai | On the free plan, images are public by default, and other users can copy and remix them. Paid plans add a private mode. |
| Adobe Firefly | Adobe says its own Firefly models are trained on licensed content, such as Adobe Stock, and public-domain material. Enterprise customers who buy the right entitlement get IP indemnification for eligible outputs; partner models inside Firefly are excluded from that Firefly cover. |
| Microsoft Copilot (commercial) | Under its Customer Copyright Commitment, Microsoft says it will defend paying commercial customers sued for copyright over output, if they used the built-in guardrails and content filters. |
Two things to remember. Ownership in the terms is a contract between you and the company: it cannot create copyright that the law does not give. And an indemnity is a promise to defend certain customers under certain conditions, not a guarantee that every output is free of other people's rights.
Famous characters, living artists and real people
Courts are also dealing with how models were trained. In June 2025, Disney and Universal sued Midjourney in a US federal court, saying it reproduced characters such as Darth Vader and the Minions. In November 2025, the High Court in London ruled mostly against Getty Images in its case against Stability AI, on the narrow claims that were left in the case.
These cases are about the companies that make the tools, not about you. But they point to a practical rule: an image that shows a famous character, another company's brand or a recognisable person can get you into trouble, whoever owns the tool. So keep living artists' names out of your prompts, get written consent before you use a real person's face or voice, and leave other companies' characters and logos alone.
Logos: trademark is a different question
A logo exists to identify a business, and that is the job of trademark law. Registration depends mainly on whether the mark is distinctive and not confusingly similar to marks already registered in the same class, country by country. Copyright in the drawing is a separate matter.
For a generated mark, the risks stack up: thin or no copyright, possible similarity to existing marks, and Arabic letters that many tools still draw badly. I go through them, with a safer workflow, in designing a logo with AI. Before filing, search the register and speak to a trademark lawyer.
How I label AI images
I direct AI tools for image, video and voice in my video work for SkillUp, and I label AI-generated images on this site. On Kababgy Al-Sultan, a grill house identity in Egypt, the takeaway and menu scenes were generated with AI around the real designs. The box, the bag, the wrap and the menu are my artwork; the settings around them are generated, and the page says so. The client has real, printable design, and the viewer knows which images are scenes.
Label in the same places your work travels: portfolios, client presentations, social posts and campaigns. The wider picture of where AI fits in a design job is in AI in graphic design.
A practical checklist
For designers
- Keep your process files. Sketches, prompts, versions and edits, with dates.
- Add real human work where it matters. Draw the mark and the lettering, compose the layout, retouch and edit. Those are the parts the law protects.
- Label AI images honestly, in your portfolio and in what you hand to clients.
- Check each tool's terms and your plan before a client project, and save a dated copy.
- Keep client secrets out of public-by-default tools. An unreleased product does not belong in a public gallery.
- Avoid living artists' names and real people's likeness, famous characters and other brands.
- Write it into the contract: which parts are generated, with which tool, and what rights pass. Do not promise exclusive copyright in a purely generated image.
For clients
- Ask which parts are generated, and with which tool.
- Ask for a short note on key assets, or for the process files.
- For a logo, ask for a mark drawn and developed by a person, with vector files, and search the trademark register before you print.
- Read what rights the contract gives you, and for anything high-stakes, ask a lawyer.
Conclusion
An AI image raises three questions, not one: is it protected, may I use it, and does it copy anyone? In the US, a prompt alone does not make you an author, while your own drawing, edits and arrangement can be protected. Elsewhere the rules are still forming, so check them where you work. Read each tool's terms, label generated images, keep your process files, and keep the parts that carry the brand, above all the logo, in human hands.
If you want an identity or a campaign where every key piece is drawn by a person and AI is used openly as a tool, tell me about your project.
Questions people ask
Are AI-generated images copyrighted?
In the United States, an image generated from a prompt alone is not protected, because copyright needs a human author. The parts a person adds can be, such as their own drawing, their edits or the way they arrange several images. Other countries are still settling the question, so check the rules where you work.
Can I use AI images commercially?
Often yes, but that permission comes from the tool's terms and your plan, not from copyright law. Read the terms on ownership, commercial use and privacy, keep famous characters, brands and real people out of the image, and do not promise a client exclusive rights to a purely generated picture.
Who owns an AI image, me or the AI company?
Many tools' terms give the output to you, OpenAI's and Midjourney's among them, with conditions that can depend on your plan. That is a contract between you and the company. It cannot create copyright where the law gives none, so others may still be free to copy a purely generated image.
Can I trademark a logo made with AI?
Trademark is a separate question from copyright. Registration depends mainly on whether the mark is distinctive and not too close to existing marks, so search the registers first and speak to a trademark lawyer before you file. A mark that a designer has redrawn and developed stands on firmer ground.
Do I have to label AI-generated images?
It depends on where the image is published and what it shows. From 2 August 2026, the EU AI Act requires deepfakes to be disclosed, with a lighter touch for clearly artistic work, and some platforms have their own labels. Labelling honestly is good practice everywhere.
