Kilde › Guides › AI Act Art 50 › Deepfakes and AI video: the 50(2)/50(4…

Deepfakes and AI video: the 50(2)/50(4) split, and the exclusions read narrowly

Current to 26 August 2026 · updates land in the changelog.

AI video engages Article 50 twice. The provider of the generating system marks the output machine-readably (50(2)); the deployer who publishes a deepfake — content resembling real persons, places or events that would falsely appear authentic — discloses the artificial generation visibly (50(4)). Different duties, different actors, and since 2 August 2026 both are live (marking with one transition for older systems; disclosure with none).

What escaped deepfake scope — less than people hope

The final guidelines exclude “wholly fantastical content, or content no audience would take as genuine”. But the boundary term — content that could plausibly exist — reads broadly, so a photoreal person saying invented words sits inside scope even in fiction-adjacent formats. The artistic, creative, satirical and fictional path is not an exemption but a softer manner of compliance: disclosure “in a manner that does not hamper enjoyment”. And where content has informative character, the guidelines say that character always prevails — advertising counts as “evidently creative” only in specific situations, with the Commission's own advertising examples all requiring standard labeling.

What deployer disclosure looks like

Visible, clear, at first exposure, repeated where viewers can join mid-content (Article 50(5) plus the guidelines) — the official EU icons exist for exactly this visible layer, though they are voluntary and icon-plus-text performs best in the Commission's own research. One boundary worth knowing: there is no retroactive duty — content published before 2 August 2026 does not need back-labeling.

The compliance file for video products

Provider side: marking implementation plus survival evidence across your delivery paths (platform re-encoding strips marks — that is the known failure mode). Deployer side: a labeling decision record per format — deepfake or not, informative or creative, standard label or non-hampering manner — so the call is documented before a regulator asks.

Related guides

Quick answers

Do deepfakes have to be labeled in the EU?
Yes — deployers must disclose that the content is artificially generated or manipulated (Article 50(4)), applying since 2 August 2026 with no transition, alongside the provider's machine-readable marking duty under 50(2).
Is artistic or satirical AI video exempt?
Not exempt — accommodated: disclosure may be delivered in a manner that does not hamper the work's enjoyment. The final guidelines interpret the creative carve-out strictly, and informative character always prevails.
Is obviously fake, fantastical video in scope?
The final guidelines exclude wholly fantastical content no audience would take as genuine, but read the boundary broadly — anything that could plausibly exist stays in scope.
Ship the disclosures before enforcement finds the gap.

The AI Act Article 50 Kit: the disclosure copy library (English + 中文), the marking implementation guide with survival-test protocol, deepfake and text-labeling walkthroughs, obviousness and scope memos, the evidence log, the Code of Practice accession path — built from the regulation and the final guidelines, with pinpoint citations.

Get the kit — US$190 Free 4-page sample (PDF)

Instant download · 14-day unconditional refund · single-organisation licence · full product page

General information only — not legal advice, and no clearance opinions. Sources are Regulation (EU) 2024/1689 (Articles 50 and 99), Regulation (EU) 2026/1744, the Commission's final Article 50 guidelines of 20 July 2026 (non-binding) and our audited kit research. Marking-technology status changes fast: treat vendor announcements as live only once verified. © 2026 Kilde.

Built by Kilde's founder, a practising attorney admitted to a US state bar (not an EU or Hong Kong admission). About · Verification log · Refunds · Terms · Privacy · esau@trykilde.com