AI labels are law now. Fashion is arguing about faces.

Welcome to the first VODOO Signal. Once a week: the AI news that changes what you can make, what you have to say about it, and one brief to copy.

This week the rules caught up with the pictures. Labelling AI content is now a legal duty in Europe, New York wants a clear notice on any ad with a synthetic person in it, and fashion is finding out what its model contracts never covered.

Rules

Europe's AI labelling duty is in force

Article 50 of the EU AI Act has applied since 2 August 2026. Whoever publishes a deepfake, meaning realistic AI images, audio or video of real-looking people, places or events, has to say it was made or changed by AI. The tools that make it have to mark their output in a way a machine can read.

Obviously creative or satirical work gets a lighter touch: the disclosure can be made in a way that doesn't spoil it.

If your ads run in the EU and show realistic people or places that never stood in front of a camera, label them. Every file made in VODOO already carries the machine-readable AI marker.

Source: EU AI Act, Regulation (EU) 2024/1689, Article 50 (EUR-Lex)

Rules

New York: an ad with a synthetic performer must say so

Since June 2026, an advertisement shown in New York that uses a synthetic performer, a digital person made by AI who looks like a real one, has to disclose it conspicuously. California has passed its own version, starting 1 January 2027.

For UGC-style and AI-influencer ads, a clear on-screen note is now the safe default in the US, not just good manners.

Source: New York State Senate, bill S8420-A

Fashion

Whose face is it? Shoot contracts meet AI

Reuters reported this week on unsigned shoot contracts that ask models to agree to their faces and bodies being altered, or swapped with other people's, using AI. Fast-fashion brands such as Mango already use AI-generated models at scale; Moncler told Reuters it does not use AI models on its site or in campaigns.

The cost argument is the reason this keeps coming up: an AI model shot can cost a fraction of a studio day. The trust argument is why luxury is holding back.

Use invented faces, or real people only with their written consent for exactly what you will make. VODOO keeps consent records for any real face or voice for this reason.

Source: FashionNetwork, reporting Reuters

New model

Kling 4.0 arrives

Kuaishou's Kling 4.0 rolled out globally in October, after an early Flash version opened to its top subscribers at the end of September. Kling has been the engine to beat for natural human motion, so this is the one to test on people-heavy shots first.

We run every new engine on the same briefs before it goes into VODOO's line-up, and we'll say here what it is best at.

Source: Digital Applied, October 2026 model release tracker

The brief · copy this

A lookbook frame that looks photographed

Full-length editorial lookbook photo of a model in a camel wool coat walking past a sunlit stone wall, mid-stride, soft late-afternoon light, natural skin texture with visible pores, slight motion in the hem, 50mm lens, shallow depth of field, muted film colour

Make this in VODOO →