If you make or use AI-generated images, you need to know three things: who (if anyone) owns them, what rights the tool’s terms give you, and when you have to tell people an image is AI-generated.
In short
- In the US, the Copyright Office says prompts alone don’t make you the author of an AI image; human-made elements and creative arrangement can still be protected. UK law has a separate, debated rule for “computer-generated works”.
- Read the tool’s terms for ownership and commercial use, and follow your university’s AI policy for coursework.
- Since 2 August 2026, the EU AI Act requires deployers to disclose deepfakes, with lighter rules for clearly artistic or creative work. Label AI images honestly wherever you publish them.
In this guide · 9 sections
Why this matters in fashion
AI image tools are now part of many fashion workflows: mood imagery, print ideas, quick visualisations of a silhouette, campaign concepts, product backgrounds and virtual models. That raises practical questions:
- If I generate an image, do I own it?
- Can I use it in my portfolio, sell it on a T-shirt or put it in a brand campaign?
- Do I have to tell people it’s AI-generated?
- What happens if it looks like someone else’s work, or a real person?
This guide summarises where the law and guidance stood in late 2026 in the US and the UK, plus the EU’s new AI labelling rules, the systems fashion students in Britain and Europe most often encounter. The law in this area is new and still changing. This is general information, not legal advice. For anything commercial or high-stakes, check the latest position and get professional advice.
Part 1: Can AI images be protected by copyright?
United States
In January 2025, the US Copyright Office published Copyright and Artificial Intelligence, Part 2: Copyrightability. Its key conclusions included:
- Copyright requires a human author. Copyright protects original expression created by a human.
- Prompts alone aren’t enough. The Office concluded that, given current generally available technology, prompts alone don’t provide sufficient human control to make users of an AI system the authors of the output.
- Human contributions can still be protected. Where a human creates expressive material, such as their own drawing that the AI modifies, or creatively selects, arranges or modifies AI-generated material, those human contributions may be protected.
- It’s case by case. Whether a particular work qualifies depends on the facts.
In March 2025, in Thaler v. Perlmutter, the US Court of Appeals for the D.C. Circuit upheld the Copyright Office’s refusal to register an artwork listed as created autonomously by an AI system, confirming that US copyright law requires a human author.
United Kingdom
UK law is different. Section 9(3) of the Copyright, Designs and Patents Act 1988 provides protection for “computer-generated works” where there is no human author. In that case, the author is taken to be the person who made the arrangements necessary for the work to be created.
How this rule applies to modern generative AI is unclear and debated. In December 2024, the UK government opened a consultation on copyright and AI, which ran until February 2025 and included questions on whether to keep, reform or remove this protection. Policy is still developing, so check the latest government position before relying on it.
| Question | US | UK |
|---|---|---|
| Is a purely prompt-generated image protected? | No, according to the Copyright Office | Possibly, under the s.9(3) computer-generated works rule; its application is debated |
| Are human-made parts protected? | Yes, the human contributions can be | Yes, as ordinary works |
| Is the law settled? | Guidance is clear; details are case by case | No; under government review |
What this means in practice
If ownership of an image matters, for example a print you want to sell or a campaign you want to protect, keep human authorship visible and documented:
- start from your own sketches, photographs or textile samples
- use AI to adapt or extend your work, not to replace it
- make significant creative edits, selections and arrangements yourself
- keep a record: your original files, prompts, iterations and edits
Part 2: Copyright in what the AI learned from
Separate from who owns the output is the question of what the AI was trained on, and whether outputs copy existing work.
Training data cases
Courts in several countries are hearing cases about AI training. In the UK, Getty Images v Stability AI was decided by the High Court in November 2025. The court dismissed Getty’s secondary copyright infringement claim, finding that the Stable Diffusion model was not an “infringing copy” because it did not store or reproduce Getty’s images. It found only limited trade mark infringement relating to watermarks appearing in some outputs. The case was narrow, and many questions about training and outputs remain open.
Outputs that look like existing work
Whatever the training position, you’re responsible for what you publish. If an AI output closely resembles a protected design, artwork, photograph, logo or character, using it could infringe someone else’s rights, just as copying it by hand would.
Practical rules:
- Don’t prompt with living artists’ or designers’ names to imitate their style for commercial work. Even where style itself isn’t protected, outputs may reproduce protected elements, and it raises ethical concerns.
- Never use AI to recreate logos, monograms or trademarked patterns.
- Check outputs for anything that looks like a recognisable existing work, a brand mark or a watermark.
Tools trained on licensed content
Some tools say they train only on content they have rights to use. Adobe, for example, says its Firefly models are trained only on data it has the right to use, such as Adobe Stock images, openly licensed content and public-domain content. That may reduce some risks for commercial use, but read the current terms and documentation yourself.
Part 3: What the tool’s terms say
Every AI tool has terms of service, and they matter as much as the law. Before using a tool for anything beyond private experiments, check:
- Who owns the outputs? Some tools assign whatever rights exist to you; others keep rights or grant you a licence.
- Is commercial use allowed? Some free tiers restrict commercial use, or make outputs public.
- What happens to your inputs? If you upload your own designs or a client’s unreleased collection, can they be used to train the model?
- Are there content rules? Many tools ban images of real people without consent, or misleading content.
- Is there any indemnity? Some business plans offer legal protection for certain uses; free plans usually don’t.
Terms change, so save a copy or note the date you checked.
Confidential work
Be careful uploading unreleased designs, client work or anything under a confidentiality agreement. Check whether the tool uses inputs for training, and whether your employer or client allows it.
Part 4: When you have to label AI images
EU AI Act: Article 50
The EU AI Act (Regulation (EU) 2024/1689) includes transparency rules in Article 50, which apply from 2 August 2026. They cover four situations:
- when AI interacts directly with people, such as chatbots
- when AI generates synthetic content, which providers must mark in a machine-readable way
- when AI is used for emotion recognition or biometric categorisation
- when AI creates deepfakes, or text published on matters of public interest
Providers (the companies that build the tools) are responsible for the first two. Deployers (the people and organisations using the tools) are responsible for the second two.
For fashion, the most relevant parts are:
- Machine-readable marking (providers): generative AI providers must mark outputs so they can be detected as AI-generated. Under the AI Omnibus agreement of May 2026, systems already on the market before 2 August 2026 have until 2 December 2026 to comply. New systems have no grace period.
- Deepfake disclosure (deployers): if you use AI to create a deepfake, meaning realistic image, audio or video content that resembles real people, objects, places or events and could falsely appear authentic, you must disclose that it’s AI-generated or manipulated. For evidently artistic, creative, satirical or fictional work, the obligation is lighter: you still need to disclose, but in a way that doesn’t hamper the display or enjoyment of the work.
The European Commission published guidelines on these obligations in July 2026. Article 50 can apply to deployers outside the EU where outputs are used in the EU, for example by posting them on the globally accessible internet.
UK and advertising rules
The UK doesn’t have an equivalent AI labelling law. However, existing rules still apply. Advertising must not mislead. If an AI-generated image in an ad gives a false impression of what a product looks like or does, for example by exaggerating its effect, that could breach advertising rules, whether or not AI was used. The Advertising Standards Authority (ASA) regulates UK advertising.
Platforms and your university
- Social platforms have their own AI labelling policies, and some apply labels automatically. Check each platform’s current rules.
- Your university almost certainly has an AI policy for coursework. UAL, for example, publishes guidance on AI and assessment, and the Russell Group has published principles on the use of generative AI in education. If you use AI in assessed work, follow your course’s rules and acknowledge it.
Part 5: Real people, models and likeness
Fashion often involves people, which raises extra issues:
- Real people: Don’t create realistic images of real people, including celebrities, without permission. Under the EU AI Act, these may be deepfakes requiring disclosure. They may also raise privacy, defamation, data protection or passing-off issues, and many tools ban them.
- Virtual models: If a brand uses AI-generated models, be transparent about it. Think about representation, and about whether the image could mislead customers about how a garment fits.
- Altering real photos: Editing a photo of a real model with AI, for example changing their body shape or skin, needs their consent, and may mislead customers.
A worked example
Worked example: a student print collection (invented)
Leila is designing a print collection for her final-year project and wants to sell a small run of scarves afterwards.
What she does:
- Paints six watercolour florals by hand and photographs them.
- Uses an AI tool to generate variations of her paintings, with her own images as inputs.
- Selects four variations, then redraws and recolours them in Photoshop, combining elements with her original paintings.
- Arranges the final repeat patterns herself.
- Keeps a process log: original paintings, every prompt, the variations generated, and her edits.
Why this helps:
- Her hand-painted work and her substantial edits and arrangements are clearly her own creative contributions.
- Her process log shows how the work was made, for her tutors and if anyone ever questions it.
What she checks:
- The tool’s terms allow commercial use of outputs on her plan, and her uploads aren’t used for training.
- Her course’s AI policy, and how to acknowledge AI use in her submission.
- None of the outputs resemble existing prints or brand patterns.
How she labels it: In her portfolio and on her shop page she notes: “Prints developed from my original watercolours, with AI-assisted variations, edited and arranged by hand.” It’s honest, and it explains her process.
Common mistakes
- Assuming you own whatever you generate. In the US, prompts alone don’t make you the author, and the tool’s terms may limit your rights.
- Not reading the terms. Commercial use, ownership and training on your inputs vary by tool and plan.
- Prompting with designers’ or artists’ names for commercial work.
- Uploading confidential designs to tools that train on inputs.
- Making realistic images of real people without consent.
- Hiding AI use in coursework, client work or advertising.
- Treating the law as settled. It’s changing quickly. Check dates on everything you read, including this guide.
Checklist
- I know whether my project needs to be protectable, and I’ve kept human authorship visible.
- I’m starting from my own work where possible, and making substantial creative edits.
- I keep a process log: originals, prompts, iterations and edits.
- I’ve read the tool’s current terms on ownership, commercial use and training on inputs.
- I’m not uploading confidential or client work without permission.
- I’ve checked outputs for resemblance to existing works, logos or watermarks.
- I don’t use real people’s likeness without consent.
- For EU audiences, I disclose deepfake content, including creative work, in an appropriate way.
- For ads, the image doesn’t mislead about the product.
- I follow my university’s AI policy and acknowledge AI use in coursework.
- I label AI use honestly in portfolios, shops and posts.
Frequently asked questions
Can I put AI-generated images in my portfolio?
Usually, yes, if you’re open about how they were made and the work shows your own skills and decisions. Many tutors and employers want to see your process, not just a polished output.
Is AI-generated art “stealing”?
That’s a live debate, not a settled legal answer. Courts are deciding cases about training, and governments are reviewing the law. What you can control is your own practice: start from your own work, avoid imitating named artists, choose tools carefully and be transparent.
Does the EU AI Act apply to me in the UK?
It can. The deployer obligations can apply where content is used in the EU, for example when it’s posted on the open internet and seen there. If you publish to EU audiences, it’s sensible to follow the disclosure rules.
What counts as a deepfake?
Under the AI Act, broadly, AI-generated or manipulated image, audio or video content that resembles existing people, objects, places or events, and would falsely appear authentic or truthful to a person. A realistic AI image of a real celebrity wearing your design would likely qualify. An obviously illustrated fantasy print would not.
Where can I check the latest position?
The US Copyright Office’s AI pages, the UK government’s copyright and AI consultation page, and the European Commission’s AI Act pages are the primary sources. Law firm briefings are useful for summaries, but check their dates.
How to cite this guide
APA 7Shreya S. (2026, November 17). AI Images in Fashion: Copyright, Ownership and Labelling. Ayerhs Magazine. https://ayerhsmagazine.com/learn/guides/ai-images-copyright-and-labelling/MLA 9Shreya S. "AI Images in Fashion: Copyright, Ownership and Labelling." Ayerhs Magazine, 17 Nov. 2026, ayerhsmagazine.com/learn/guides/ai-images-copyright-and-labelling/.Harvard (check your university’s version)Shreya S. (2026) AI Images in Fashion: Copyright, Ownership and Labelling. Ayerhs Magazine. Available at: https://ayerhsmagazine.com/learn/guides/ai-images-copyright-and-labelling/ (Accessed: [date you read it]).
Related on Ayerhs: How to use AI in the design process · How to use AI in coursework honestly · How to work with fashion influencers
Sources. Checked 17 November 2026. Formats and policies change, so always follow your own course handbook and tutor first.
- US Copyright Office. Copyright and Artificial Intelligence, Part 2: Copyrightability (29 January 2025). copyright.gov
- Sidley Austin. “U.S. Copyright Office issues report on artificial intelligence and copyrightability” (7 February 2025). datamatters.sidley.com
- Ballard Spahr. “Copyright Office issues report on copyrightability of AI content” (31 January 2025). ballardspahr.com
- US Court of Appeals for the D.C. Circuit. Thaler v. Perlmutter, No. 23-5233 (18 March 2025). cadc.uscourts.gov
- Copyright, Designs and Patents Act 1988, section 9(3). legislation.gov.uk
- UK Government (DSIT, DCMS and IPO). “Copyright and Artificial Intelligence” consultation (17 December 2024 to 25 February 2025). gov.uk
- Getty Images (US) Inc and others v Stability AI Ltd [2025] EWHC 2863 (Ch), judgment of 4 November 2025. judiciary.uk
- Slaughter and May. “Mining the copyright aspects of Getty Images v Stability AI” (2025). slaughterandmay.com
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 50, via the AI Act Explorer. artificialintelligenceact.eu
- Future of Life Institute, AI Act resources. “The EU AI Act’s transparency rules: a practical guide to Article 50”. artificialintelligenceact.eu
- Travers Smith. “Is it a bot? EU AI Act transparency rules take effect 2 August 2026”. traverssmith.com
- Ropes & Gray. “You talkin’ to me? Operationalising the EU AI Act’s transparency obligations”. ropesgray.com
- Addleshaw Goddard. Briefing on the European Commission’s Article 50 guidelines. addleshawgoddard.com
- Adobe. “Adobe Firefly FAQ” (training data). helpx.adobe.com
- Advertising Standards Authority. “About the ASA and CAP”. asa.org.uk
- University of the Arts London. “AI and assessment”. arts.ac.uk
- Russell Group. “Russell Group principles on the use of generative AI tools in education” (July 2023). russellgroup.ac.uk
← All guides · Learn

