Do AI Marketing Videos Need a Label? EU AI Act Article 50 for Founders (November 2026)
Sometimes. Since August 2, 2026, Article 50 of the EU AI Act requires businesses that publish AI-generated video that counts as a 'deep fake' to disclose it. That covers realistic content that resembles real people, places, or events and could pass as authentic. In practice, a photorealistic AI avatar or AI UGC creator in an ad should carry a visible AI label for EU audiences. Designed motion graphics, such as kinetic type, animated UI, and illustrated scenes, generally aren't deep fakes, so the disclosure duty usually doesn't apply to them. Separately, the companies that make AI video tools must mark their output in a machine-readable way. This is a plain-English summary, not legal advice.
Two duties: the tool's and yours
| Provider duty (Art. 50(2)) | Deployer duty (Art. 50(4)) | |
|---|---|---|
| Who | The company that makes the AI tool | The business that uses it and publishes the content: you |
| What | Mark synthetic audio, image, video, and text in a machine-readable, detectable way | Disclose that deep-fake content is artificially generated or manipulated |
| Since | August 2, 2026; tools on the market before then have until December 2, 2026 | August 2, 2026, with no grace period |
| Exceptions | Assistive or standard editing that doesn't substantially alter the input | Evidently artistic, creative, satirical, or fictional works get a lighter, non-intrusive disclosure |
Sources: the Commission's AI Act Service Desk text of Article 50 and the transitional rule added by the Digital Omnibus, Regulation (EU) 2026/1744, in Article 111. That rule gives providers of systems placed on the market before August 2, 2026 until December 2, 2026 to comply with the marking duty. It does not delay the deployer's disclosure duty.
What counts as a deep fake?
Article 3(60) defines a deep fake as AI-generated or manipulated image, audio, or video content that resembles existing persons, objects, places, entities, or events and would falsely appear to a person to be authentic or truthful. The key test is whether a reasonable viewer could take it as real. Law-firm summaries of the Commission's final Article 50 guidelines (published July 20, 2026) report that content needn't depict an actual person to qualify, so an invented but photorealistic spokesperson is likely in scope. Those guidelines aren't binding, but national regulators are expected to follow them.
Which AI marketing videos need a label?
| Video type | Likely a deep fake? | What to do for EU audiences |
|---|---|---|
| Photoreal AI avatar or AI UGC creator | Yes, likely | Visible AI label from the first moment, such as an on-screen tag or caption |
| Your own face or voice cloned by AI | Can be, if it shows you saying things you didn't film | Label it; don't rely on the fact that it's you |
| Realistic generated b-roll (people, places, events) | Can be | Label if a viewer could take it as real footage |
| Motion graphics: kinetic type, animated real UI, illustration | Generally no | No deep-fake disclosure needed; normal ad rules still apply |
| Cartoon or clearly stylized characters | Generally no | Usually no disclosure needed |
| AI-written script read by a real filmed person | No (video isn't synthetic) | No Article 50 video disclosure |
Commentary on the guidelines from firms including KPMG Law and McCann FitzGerald says the disclosure should be noticeable at first exposure and that a machine-readable mark alone isn't enough for deep fakes. The creative-work exception is read narrowly, and commentators doubt it covers ordinary promotional content.
Does this apply to a US startup?
Yes, if the output is used in the EU. Article 2 covers providers and deployers located outside the EU where the AI system's output is used in the Union. A US company running an AI-avatar ad aimed at European audiences is a deployer under the Act.
Penalties
Breaches of Article 50 can be fined up to EUR 15 million or 3% of worldwide annual turnover, whichever is higher, under Article 99(4). For SMEs and start-ups, the cap is whichever of the two is lower. Fines are set and enforced by national authorities.
Platform labels apply on top of the law
- YouTube requires creators to disclose realistic AI-generated or altered content, including making a real person appear to say or do something they didn't, through the AI-use question in YouTube Studio. AI-written scripts and unrealistic content don't need it. Source: YouTube Help.
- TikTok requires labels on AI-generated content showing realistic-looking scenes or people, may label it itself, and auto-labels content carrying C2PA Content Credentials.
- Meta labels ads containing AI-generated photorealistic people made with its own tools, and shows 'AI info' in About this ad when it detects third-party AI signals. Organic photorealistic AI video must use Meta's disclosure tool.
- Ad-specific rules on each platform can differ from organic rules; check the ads policy for the account you're running.
The US angle: fake testimonials
In the US, the FTC's rule on consumer reviews and testimonials (16 CFR Part 465, in effect since October 2024) bans testimonials that misrepresent that they come from someone who doesn't exist, such as AI-generated fake reviews, or from someone without real experience of the product. An AI avatar presented as a happy customer is high-risk even if labeled. An AI spokesperson who clearly isn't posing as a customer is a different case. See the FTC announcement.
A practical checklist for founders
- Inventory which ads use realistic AI people, voices, or scenes. Those are your deep-fake candidates.
- Add a visible label to them for EU audiences, such as 'AI-generated' on screen from the first frame or in the caption where the platform shows it immediately.
- Use each platform's own AI disclosure toggle as well.
- Never script an AI actor as a real customer. Use 'spokesperson' framing, or real customers with consent.
- Ask your AI video vendor how it marks output (for example, C2PA Content Credentials) and whether it offers a built-in label. Our Arcads alternatives comparison notes which tools document this.
- Where the product itself is the message, consider motion graphics showing your real UI. It's usually outside deep-fake territory and is clearer for software anyway; see motion graphics vs text-to-video.
The Commission also published a voluntary Code of Practice on marking and labelling AI-generated content on June 10, 2026, including an optional EU icon for labels. Following it isn't required, but it's the clearest signal of what regulators consider good practice.
AI video labels FAQ
Does a motion graphics video made with AI need an AI label?
Generally not under Article 50(4). Kinetic type, animated real UI, and illustrated scenes don't falsely appear authentic, so they aren't deep fakes. The tool's provider still has a separate machine-readable marking duty.
Do AI avatar ads need a label in the EU?
Most likely, yes. A photorealistic AI person that a viewer could take as real is a deep fake under Article 3(60), even if the person is invented, and the deployer must disclose it.
Was Article 50 delayed by the Digital Omnibus?
No. It has applied since August 2, 2026. Regulation (EU) 2026/1744 only gave providers of AI systems already on the market before that date until December 2, 2026 to meet the machine-readable marking duty.
What is machine-readable marking?
A signal embedded in the file, such as C2PA Content Credentials, metadata, or a watermark, that lets software detect content as AI-generated. It's the tool provider's job, and it doesn't replace a visible label for deep fakes.
Does the EU AI Act apply to US startups?
Yes, when the AI output is used in the EU. Article 2 covers providers and deployers outside the Union in that case.
Is a cloned version of my own voice or face a deep fake?
It can be, if it shows you saying or doing something you didn't actually film or record, and a viewer could take it as real. Labeling it is the safe default.
Last checked November 12, 2026 against the Commission's AI Act Service Desk text. This guide is general information, not legal advice; for a specific campaign, ask a lawyer familiar with the AI Act.
Make your next launch in motion
Bring a brief, brand assets, and product captures. Review the story, visuals, and exported video before publishing. Request free private-beta access.
Get early access →