The EU's labelling duty for generated video just started to bind

The figure or the record

The EU's Artificial Intelligence Act, Regulation (EU) 2024/1689, was adopted after the European Parliament's position of 13 March 2024 and the Council's decision of 21 May 2024, and was done at Brussels on 13 June 2024. Article 113 separates adoption from application in its own text: the Regulation entered into force twenty days after publication, but 'it shall apply from 2 August 2026', the date this record uses as the event, with narrower carve-outs applying earlier (prohibited-practice rules from 2 February 2025) and later (certain high-risk obligations from 2 August 2027). Article 50, the transparency article covering generated content, falls under the general 2 August 2026 date, meaning its duties became applicable one month before this record was prepared.

What the documents establish

Article 50(2) requires providers of AI systems, including general-purpose ones, that generate synthetic audio, image, video or text to mark the output in a machine-readable format detectable as artificially generated or manipulated, so far as technically feasible. Article 50(4) separately requires a deployer whose system generates or manipulates a deepfake to disclose that the content is artificially generated or manipulated, with an explicit carve-out: where the content is part of 'an evidently artistic, creative, satirical, fictional or analogous work', the duty narrows to disclosing the existence of the generated content in a way that does not hamper displaying or enjoying the work, rather than a disclaimer overlaid on it.

Scope and caveats

Article 50 imposes a labelling and disclosure duty; it does not require consent from the person depicted, which remains a matter for national likeness law such as Denmark's proposed Copyright Act changes. The Commission's own AI Act page, a living document reflecting the site as retrieved 16 September 2026, confirms in plain terms that 'the transparency rules of the AI Act will come into effect in August 2026', without adding detail the Regulation's text does not already state. The duty runs against providers and deployers placing systems on the EU market; it does not dictate how a non-EU production must label content it never distributes into the Union.

What to watch

Editorially, the test worth watching is how narrowly regulators read 'evidently artistic, creative, satirical, fictional or analogous', since a vertical drama using an AI-generated performer is squarely fictional entertainment, and whether that exemption is read to require only an unobtrusive existence-disclosure, rather than an on-screen deepfake warning, will determine how the labelling duty actually reads on a finished episode.

• Is the AI-generated content machine-readably marked under Article 50(2), separately from any human-facing disclosure under Article 50(4)?
• Does the production's fictional framing qualify for the narrower 'existence disclosure' standard, or does it risk creating the false impression of an authentic recording the exemption does not protect?
• Is the obligation being assessed against the provider (the tool maker) or the deployer (the production), since Article 50 assigns different duties to each?

Article 50 took effect only weeks before this record was prepared, so any enforcement practice cited against it should be checked for whether it predates 2 August 2026, when the duty was not yet applicable at all.