Guide · Art. 50(4)

AI Deepfake Disclosure Requirements — Article 50(4)

Last updated: 8 November 2026

AI deepfake disclosure requirements under Article 50(4) fall on the deployer: realistic AI-generated or manipulated image, audio or video of real people, places or events must be disclosed as artificially generated or manipulated. Evidently artistic or satirical work needs only a lighter disclosure. The free Article50.io scan shows the most severe finding on one public page.

Article 50(4) of the EU AI Act covers two things: deep fakes, and AI-generated text published to inform the public. In both cases the deployer must disclose that the content was artificially generated or manipulated.

Scan for free

What counts as a deep fake

Article 3(60) of the AI Act 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.

That's broader than face-swapped videos of politicians. A synthetic "photo" of your real office, a cloned voice of your CEO in a product video, or an AI-edited image of a real event can all qualify — if a viewer could mistake it for the real thing.

Image, audio or video

The definition covers these three media. Text is handled separately, below.

Resembles something real

Existing people, objects, places, entities or events. Paragraph 113 (C(2026) 7702 final, 5 November 2026, non-binding) says this can include a subject that exists, can plausibly exist, or could plausibly have existed. That is one criterion, not the whole definition. A photorealistic invented person is not automatically in scope. The 20 July 2026 text was a draft and said the same.

Could pass as authentic

It would falsely appear authentic or truthful. Paragraph 114 says this is assessed in context, together with resemblance. A high degree of photorealism makes that more likely, but photorealism alone is not determinative.

Who must disclose

Article 50(4) puts the obligation on deployers: whoever uses the AI system to generate or manipulate the content in a professional capacity. If your marketing team generates a realistic image with an AI tool and publishes it on your site, the disclosure is yours to make — separate from the tool maker's own Article 50(2) marking obligation.

The disclosure must state that the content has been artificially generated or manipulated, clearly and at the latest when the viewer is first exposed to it. For evidently artistic, creative, satirical or fictional work, a lighter disclosure applies: flag that AI-generated content is present, in a way that doesn't spoil the work. Uses authorised by law to detect, prevent, investigate or prosecute criminal offences are exempt.

The public-interest text disclosure

The same sub-clause covers text. Deployers of an AI system that generates or manipulates text published with the purpose of informing the public on matters of public interest must disclose that the text was artificially generated or manipulated.

The exemption: text that has undergone human review or editorial control, where a natural or legal person holds editorial responsibility for its publication. A newsroom where an editor signs off on AI-assisted copy is in a different position from a site auto-publishing AI-written news summaries.

Image caption

<figcaption class="ai-label">
  AI-generated image. This picture was created with artificial
  intelligence and does not show a real [person / place / event].
</figcaption>

Deep fake disclosure checklist

✔Realistic AI-generated or AI-edited images of real people, places or events are labeled
✔AI voice clones and synthetic video carry an on-screen or in-player disclosure
✔Labels sit on or directly beside the content, not on a separate page
✔AI-written public-interest articles are disclosed — or genuinely edited by someone who holds editorial responsibility

See how this fits with the other obligations in Article 50 of the EU AI Act, explained, or run through the full compliance checklist.

Frequently Asked Questions

What counts as a deep fake under the EU AI Act?

Article 3(60) of the AI Act 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.

Who has to disclose a deep fake?

Under Article 50(4), the deployer — the person or organisation using the AI system to generate or manipulate the content in a professional capacity. If you publish it on your website, that's usually you.

Do satirical or artistic deep fakes need a disclosure?

Yes, but a lighter one. Where the content is part of an evidently artistic, creative, satirical, fictional or analogous work, the obligation is limited to disclosing the existence of the generated or manipulated content in an appropriate way that doesn't hamper the display or enjoyment of the work.

Does AI-written text count as a deep fake?

No — the deep fake definition covers image, audio and video. AI-generated text is handled by the second half of Article 50(4): it must be disclosed when published to inform the public on matters of public interest, unless it has undergone human review or editorial control and someone holds editorial responsibility for it.

Check your site automatically

Article50.io is an automated scanner for the Article 50 transparency rules of the EU AI Act. It scans a website for potential Article 50 transparency gaps and generates a report of what the scan found and what it could not check, with remediation guidance, implementation instructions and disclosure wording you can paste in.

The free scan shows your single most severe finding. Scans are capped at 45 seconds — no signup, public pages only.