The ASA's position: same Codes, no new disclosure rule
The ASA and CAP have said the Codes are media-neutral: an ad is judged on the impression it gives, not on how it was made. There is no general duty to say an ad used AI. The test is whether people would be misled without that information, and a disclosure cannot rescue an ad that misleads anyway. CAP's own example is a generated image that overstates what a cosmetic does; adding 'AI-generated' does not fix it. The same logic runs through the ASA's earlier rulings on beauty filters. In June 2026 CAP added that a deepfake endorsement real enough to mislead is likely to breach the Code, and that the advertiser remains responsible for any biased output it uses. CAP's rule on harmful gender stereotypes, in force since 2019, applies to a generated cast exactly as to a real one.
A synthetic person still has an age
Several UK rules turn on how old someone seems, and the ASA applies them to generated people as readily as to cast ones. In alcohol ads, nobody shown drinking or playing a significant role may be, or seem to be, under 25; gambling ads carry a similar test. Generation models tend towards smooth, youthful faces, so a character briefed as thirty can read as under 25 on a phone screen, and that impression is what the ASA judges. Since 5 January 2026, ads for identifiable less healthy food and drink have been barred from TV and on-demand services between 5.30am and 9pm, and from paid online media at any time. Brand ads that show no such product are outside it, but a generated lifestyle scene with a recognisable product in shot is not.
Likeness without an image right
English law has no general image right. Protection comes from passing off, data protection and contract. The leading case is close to home for generated content: in Irvine v Talksport (2002) a racing driver won because a doctored photograph implied he endorsed a radio station. A synthetic lookalike or soundalike that suggests a real person's endorsement carries the same risk, and UK GDPR covers the recordings used to build any digital twin. The government's March 2026 report on copyright and AI said it would consider a new personality right over image and voice. Until one exists, permission is the only safe basis. Get it in writing, and cover the voice as well as the face, the territories and how long the assets may run.
Copyright in AI output, and TV clearance
The UK is unusual in protecting computer-generated works. Section 9(3) of the Copyright, Designs and Patents Act 1988 makes the author the person who made the arrangements for the work, with protection for 50 years. In March 2026 the government said it would prefer to remove the provision but would keep it under review, so treat ownership of purely generated material as uncertain, record the human choices behind each asset, and take rights by contract from every supplier. TV ads are cleared by Clearcast against the BCAP Code before broadcast, and radio ads by Radiocentre. Synthetic White keeps a per-asset record of which model made what, which helps when a clearance body or a complainant asks how an image was produced.
Updated 25 September 2026 · General information, not legal advice. Rules change, so check the current text with your legal team before relying on it.