Is it safe to use AI-generated images in client work?
It can be, with checks on the tool's commercial-use terms, resemblance to existing work, real people's likenesses, trade marks and misleading claims.
The main risks
AI imagery is not inherently unsafe, but it introduces risks that stock libraries and commissioned shoots usually handle for you.
Unclear ownership: You may not hold the rights you assume in a generated image, which matters if a client expects exclusivity. Who owns the copyright in AI-generated work explains why.
Resemblance to existing work: An output can come out close to a recognisable photograph or illustration, especially when a prompt names a specific artist or famous image.
Real people's likenesses: Images that look like a celebrity, influencer or identifiable member of the public create endorsement and reputational risk.
Trade marks: Logos, packaging and branded products can appear in generated scenes, sometimes distorted, sometimes clearly identifiable.
Misleading content: An image that shows a product doing more than it does, or a venue looking better than it is, can breach advertising rules however it was made.
The legal position is still developing
The UK High Court rejected Getty Images' copyright infringement claim against Stability AI. Getty's appeal is due to be heard in November 2026, so the position may change. UK copyright law on AI-generated work is unsettled more broadly, so treat today's view as provisional and avoid commitments that assume it will stay put.
Choosing tools for client work
AI image tools differ in whether their terms allow commercial use and whether the provider offers protection against intellectual property claims. Before approving a tool, check:
Commercial-use rights: Whether your plan allows commercial use at all, and any limits on use in logos, packaging or products for resale.
IP protection: Whether the provider will defend or compensate you if an output infringes someone's rights, and what you must do for that protection to apply.
Plan level: Whether these terms apply to every plan or only to business accounts.
Record what you checked and when, because terms change.
Before an image goes live
Build a short check into your sign-off process:
Reverse image search: Check whether the output closely resembles an existing image.
Faces: Confirm no one looks like a real, identifiable person unless you have permission and a clear basis for it.
Logos and text: Zoom in on backgrounds, clothing and products for stray branding or garbled lettering.
Accuracy: Ask whether the image shows anything a consumer could rely on, such as product results, food portions or event facilities. If so, it must be accurate.
Prompt record: Keep the prompt, tool category, date and edits made, in case questions arise later.
Advertising rules apply in full
The UK Advertising Codes apply to AI-generated ads in the same way as any other ad. AI-generated depictions of real people must not mislead consumers into believing there is a genuine endorsement. A video ad featuring a synthetic version of a sports personality, or a social post showing a lookalike influencer holding the product, is high risk even with a label. Labelling helps transparency but does not make a misleading image acceptable, as whether to disclose AI use explains.
Where AI imagery fits best
Lower-risk uses include mood boards, pitch concept visuals, background textures, storyboards and internal presentations. Higher-risk uses include hero campaign images, packaging, anything depicting real people, and brand assets the client needs to own outright. Copyright and advertising rules depend on the details, so take advice on your own situation.
More questions