Aipedia: Marketing and PR agencies

AI-generated images in ads: copyright, trade mark and ASA risk

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

AI-generated ad images are treated like any other ad image: they must not mislead, and they must not infringe someone else's copyright or trade mark. Agencies face both risks because they create and supply the visuals clients rely on.

Part of: AI in Marketing and PR agencies

In short

AI image generators let agencies produce campaign visuals fast, but the ASA applies its media-neutral rules to AI ads and CAP guidance says altered images must not exaggerate what a product can do. Getty Images v Stability AI [2025] EWHC 2863 (Ch) shows the copyright, trade mark and passing off arguments that generated images can attract. The IPA/ISBA principles expect agencies to consider IP holders, check permissions and agree AI use with clients. Agency PI may include some IP cover, but trade mark, contractual indemnity and campaign withdrawal costs may fall outside it.

How agencies use AI to create campaign visuals

Image generators can be used beyond mood boards for finished work. Agencies may use them to produce hero images, product scenes, backgrounds, packaging mock-ups and social assets, often in hours rather than the days a shoot would take. Some teams generate a base image and retouch it; others use AI to extend, relight or clean up photographs supplied by the client. The common thread is that the agency no longer holds a simple licence from a photographer or picture library for every element of the final image, and may not know what the tool drew on to produce it.

The rules and principles that apply

There is no AI-specific rule in the CAP Code. The ASA's 2025 Annual Report confirms its media-neutral rules apply to ads made with AI and records a rise in complaints about AI-generated ad content. CAP's guidance on misleading imagery (11 September 2025) says images altered by production techniques, including filters, must not exaggerate what a product can do. It does not mention AI, but an AI-polished image of skin, food, a room or a car is judged on the same basis.

The IPA and ISBA principles add industry expectations: consider IP holders (principle 7), carry out due diligence on tools (principle 9), keep human oversight including checking permissions (principle 10), and do not put AI content into client materials without agreement (principle 11).

What Getty v Stability AI shows about IP risk

Getty Images (US) Inc & Ors v Stability AI Ltd [2025] EWHC 2863 (Ch), decided by Joanna Smith J on 4 November 2025, is a significant UK case on image generators. Getty abandoned its claims about training and outputs because there was no evidence the training took place in the UK. The trial went ahead on secondary copyright infringement (whether the model itself is an "infringing copy"), trade mark infringement based on Getty watermarks appearing in generated images, and passing off. For agencies the lesson is practical: generated images can carry recognisable marks, styles or elements that invite infringement and passing off arguments, and the law on AI and copyright is still developing. The government's report and impact assessment on copyright and AI was published on 18 March 2026, and its consultation feedback is still being analysed.

What can go wrong

Illustrative scenario: an agency creates a campaign for a skincare client using AI-generated models with flawless skin, presented as results users can expect. A competitor complains to the ASA, which finds the imagery exaggerates the product's effect. Separately, a photographer says one background closely reproduces her published work. The client pulls the campaign, pays for a reshoot and asks the agency to cover the wasted media spend and the photographer's claim.

Other possible problems include brand logos or watermarks surfacing in generated images, a generated person resembling a real individual, and stock-style AI images reused across clients so that two brands end up with near-identical visuals.

How PI and media cover usually respond

No statutory or professional-body PI requirement applies to agencies, so cover is what you choose to buy. Agency PI may be written to include some intellectual property and media liability, and may respond to an allegation of unintentional infringement, subject to the policy terms. Wordings vary: some exclude trade mark infringement or patents, some limit IP cover to defence costs, and some exclude liability you have taken on under a contractual indemnity. A finding that an ad was misleading is not itself a claim for damages, so the cost of pulling a campaign may sit with the agency unless the client sues. Our guide to cyber and PI overlap for AI-generated content explains where IP claims tend to land, and PI or media liability for AI claims compares the policies.

Practical controls

Use generators whose terms you have read, and record which tool produced each asset.
Ban prompts naming living artists, photographers, brands or real people unless you hold rights.
Check outputs for logos, watermarks, text fragments and lookalike faces before sign-off.
Do not use AI to enhance product performance shots; keep before-and-after images genuine.
Get written client agreement to AI imagery in each campaign, as principle 11 expects.
Keep prompts, versions and approvals in case you need to show how an image was made.

Logos and brand marks raise their own clearance questions, covered in AI brand names, logos and trade mark clashes. For the insurance side of infringement more broadly, see AI IP infringement insurance and our glossary entry on IP infringement cover. The marketing and PR agencies hub lists every guide in this series.

How Apex helps: Apex arranges PI for agencies producing AI visuals. A broker reviews how your wording treats IP claims from generated imagery and any contractual indemnities you give.

Sources

Frequently asked

Can I be sued for an AI-generated image my agency created?

Yes. If an image reproduces a substantial part of a protected work, uses a trade mark, or suggests a false connection with a brand or person, the rights holder may claim against whoever published it, which can include the agency and the client. The tool provider's terms may not protect you.

Did Getty v Stability AI decide AI images are lawful?

The case did not decide training or output claims, which Getty abandoned because the training was not shown to have happened in the UK. The trial focused on secondary copyright infringement, trade mark use of watermarks and passing off. Agencies should not treat it as a general clearance for AI imagery.

Does the ASA treat AI images differently from photographs?

No. The ASA applies the same media-neutral rules. An AI image that exaggerates what a product can do is assessed in the same way as an over-filtered photograph, and the ASA reports a growing number of complaints about AI-generated ad content.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.