The rule in one paragraph
On 25 September 2025 the UAE Media Council stated that using AI or other modern tools "to depict national symbols or public figures without prior official approval constitutes a clear legal violation of media content standards", and that using AI to spread misinformation, incite hatred, defame people or attack social values is a media offence subject to fines and administrative penalties. Its successor, the National Media Authority (NMA), repeated the warning on 29 September 2026: "Not everything that can be generated is permitted."
There is no separate AI law behind this. The statements apply the existing content standards of the Media Law (respect for the state, its symbols and institutions; national identity; privacy; no false news) to a new tool. That matters, because it means the existing fines apply.
What the fines look like
From Cabinet Resolution 42/2025:
| What the AI content does | Standard | Fine range, AED |
|---|---|---|
| Depicts the system of government, its symbols or leaders disrespectfully | Governance and symbols | 50,000 – 500,000 |
| Mocks national dress, dialect or heritage | Heritage and identity | 25,000 – 150,000 |
| Fakes a news event or attributes words to a real person | False news / rumours | 10,000 – 100,000 |
| Uses a real person's likeness without consent | Privacy | 20,000 – 100,000 |
| Ad that misleads about the product (Art. 33) | Advertising conditions | 20,000, then 50,000 |
On top of media fines, a deepfake that edits a real person's image to harm them can fall under Article 44 of the Cybercrime Law: at least one year's detention and AED 250,000 to AED 500,000.
A use-case table for brands
Our reading of where common AI uses land. Not legal advice; the regulator has not published a list.
| AI use in marketing | Status | Notes |
|---|---|---|
| UAE flag, emblem, rulers, national landmarks as symbols | Prior approval | Expressly covered by the NMA statement |
| A real public figure (ruler, minister, celebrity, influencer) | Prior approval + consent | Also a privacy and likeness issue |
| Fictional AI model wearing a kandura in an ad | Avoid | NMA has said the kandura is reserved for Emiratis in commercial content |
| AI voice imitating the Emirati dialect for laughs | Avoid | NMA objected to non-Emiratis using broken Emirati dialect to sell |
| Fictional AI model in neutral clothing | Allowed | Content standards and platform labels apply |
| AI background replacement, sky, cleanup | Allowed | Fine if the product itself is shown truthfully |
| AI-rendered product or dish that looks better than reality | Risky | Art. 33 bans misleading claims and imitated images |
| AI "customer testimonial" with a synthetic face | Avoid | Presents a fake person as a real customer: misleading |
| AI-generated real estate interiors for an off-plan project | Allowed with care | Label as render; Trakheesi rules still apply |
| AI script, captions, translation | Allowed | No visual risk; check facts and Arabic quality |
Platform labels are a separate obligation
UAE law, as far as we could find, has no general duty to label AI content. The platforms do:
- TikTok requires creators to label AI-generated content that contains realistic images, audio or video, and may remove unlabeled realistic AI content. It auto-labels files carrying C2PA Content Credentials.
- Meta said in February 2024 it requires disclosure for organic posts with photorealistic video or realistic-sounding audio that was digitally created or altered, and may penalise those who don't.
A paid AI ad therefore can need two labels: "Ad" for UAE law and "AI-generated" for the platform. See our ad disclosure guide.
What nobody tells you
- "Prior approval" has no published form. We did not find an NMA service page for approving AI depictions of symbols or public figures. In practice that means: if you can't get approval, don't publish it.
- National Day is the peak risk. The 2026 restatement came weeks before 2 December. Flag-coloured AI art of leaders is exactly what the rule targets.
- The kandura rule applies to real people too. It is not an AI rule. AI just makes it easier to break by accident, because a prompt like "a businessman in Dubai" can easily come back in a kandura.
- Synthetic testimonials are the quiet trap. They don't touch symbols or celebrities, so teams assume they are safe. They present a person who doesn't exist as a customer, which is the definition of a misleading ad.
Vladimir Chernyshev, GM: use AI for cleanup, backgrounds and translations, and shoot people and products for real. In the UAE a real half-day shoot costs less than explaining an AI image to a regulator.
Related: UAE media content standards, media law fines, photographing people.
