Deepfake labeling requirements under the EU AI Act

What Article 50 says about deepfakes

Article 50 of the EU AI Act places specific obligations on deployers of deepfakes. A deepfake is AI-generated or manipulated content that resembles real persons, places, or events and would appear authentic to a reasonably well-informed person.

Deployers must disclose that the content is artificially generated or manipulated.

Who is responsible?

• Providers of AI systems that can generate deepfakes must ensure outputs are marked in a machine-readable way and detectable as artificially generated
• Deployers who publish deepfakes must disclose that the content is artificially generated or manipulated

If your company uses AI to create or manipulate video, images, or audio that resembles real people or events and publishes it, you are a deployer and you must disclose.

What wording to use

Deepfake video

"This video has been artificially generated and may not depict real events."

Deepfake image

"This image has been artificially generated and may not depict real events."

Deepfake audio (voice cloning)

"This audio has been artificially generated using AI voice synthesis and may not represent the real voice of the person depicted."

Avatar spokesperson

"This video features an AI-generated avatar and does not depict a real person."

Where to place the disclosure

• Video: At the start of the video and in the description
• Images: As a visible caption or watermark, and in alt text
• Audio: Spoken at the beginning and in the content description
• Social media: In the caption and, where possible, on the media itself

When disclosure is not required

• Personal, non-professional use — deepfakes created for personal use and not published
• Content obvious as AI-generated — if a reasonably well-informed person can tell it's not real
• Artistic or creative content — disclosure must not hamper the enjoyment of the work, but some form of disclosure is still expected
• Law enforcement, justice, or democratic processes — specific exemptions apply

Practical steps

1. Audit what deepfake or synthetic media you publish
2. Determine which content resembles real persons, places, or events
3. Apply the appropriate disclosure wording
4. Place the disclosure prominently — at the start of video/audio, on or near images
5. Log what was labeled, when, and by whom

The full deepfake notice library

The Labeling Kit (€99) includes deepfake/synthetic media notice templates for video, images, audio, and avatar spokespersons — plus a labeling guide, decision tree, and record log.