Law explained

“Deepfake” doesn't mean malicious: when AI content needs a visible label

The AI Act’s definition of a deepfake is wider than the scandal headlines: AI-generated or manipulated image, audio or video that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful (Art. 3(60)). Note what’s missing: intent. A flattering AI video of your own storefront qualifies.

The duty (Art. 50(4))

Deployers must disclose that such content has been artificially generated or manipulated. Visibly — this is about human perception, not just metadata. Clear, at first exposure (50(5)).

Where small businesses meet this rule

An AI avatar of a real employee greeting customers; AI voice-over cloned from a real voice; “photos” of your real premises generated instead of shot; AI-reconstructed event footage. All labelable.

The art/satire soft-landing

Where content is evidently artistic, creative, satirical or fictional, the duty limits itself to a disclosure that doesn’t hamper the enjoyment of the work — e.g. a credit line rather than a watermark across the art.

Your own face is still a real face

Using AI on yourself or your own premises doesn’t exempt you — authenticity-confusion is the test, not consent.

Labels that work

“AI-generated video” in the corner or caption, spoken or on-screen note for audio/video intros. Snippets and per-platform templates are in the Kit.

Sources are cited in place; official text: Regulation (EU) 2024/1689. This guide is a technical explanation, not legal advice.