Do you have to disclose AI in a UK commercial?

In May 2025 the ASA and CAP set out their view. The CAP and BCAP Codes have no AI-specific rules, and there is no blanket legal requirement in the UK to disclose AI in ads. The existing rules apply however an ad was made, edited or targeted.

Their test is the ordinary one: would the audience be misled if the AI were not disclosed, and would a disclosure correct that or contradict the ad? A label cannot rescue a misleading ad. They also say they will change their approach if the technology makes that proportionate, so treat this as current, not settled.

The ASA notes that advertisers have altered picture in post for years without a duty to say so, and that much AI use is the same kind of work.

The IAB AI Transparency and Disclosure Framework of August 2026 draws the line in a similar place. The IAB is a US trade body and the framework is voluntary. It puts routine post-production below its disclosure threshold, naming colour grading, stabilisation, background clean-up and routine visual effects, with removing safety rigging from a stunt as one example. Its rule of thumb: if an edit would not need disclosing when done with traditional tools, doing it with AI does not change that, unless the edit itself could mislead.

A reframe changes the crop, not what happened in front of the camera. Clean-up, rig removal, grading and reframes sit on this side of the line.

The same framework lists the uses that do call for a visible label: video generated by AI from a text description or a source image; photorealistic synthetic people in leading roles; a living person's voice or digital double shown saying or doing things they never did; and any digital double of a person who has died, even with the estate's agreement.

Each changes who the viewer thinks is on screen, or what they think happened. Halation does not synthesise or alter a human performance on a client production, so our work stays on the other side of this line.

Article 50(4) of the EU AI Act requires anyone using AI to produce a deep fake to disclose that the content was artificially generated or manipulated. A deep fake is image, audio or video that resembles existing people, objects, places, entities or events and would falsely appear authentic. For evidently artistic, creative or fictional work, the duty is to disclose in a way that does not spoil the work.

The IAB framework records that these obligations took effect on 2 August 2026, and that Article 50 has no carve-out for paid endorsements: an authorised voice clone of a living person still needs disclosing. How the creative-work exception applies to advertising is a question for your lawyers.

In September 2026 UK Screen Alliance and Animation UK announced a partnership with AIMICI, a UK responsible-AI organisation, offering members its consultancy and discounted ScreenSkills-accredited training. The announcement notes that studios, commissioners and clients increasingly ask how AI was used on a piece of work, and that teams need a consistent way to record and disclose it.

The rules will keep moving. The answer on any one job depends on facts: which shots AI touched, what it did, and whether any performance or likeness was involved.

Every Halation delivery carries an AI Usage Schedule that records exactly that, shot by shot, along with what was not done. The agency, the client's lawyers or a broadcaster's compliance team can then make the disclosure call for each market from a written record. If the rules tighten after delivery, the record already exists. More on the two schedules.