AI actors in brand films: what to disclose

A synthetic actor never cancels a shoot, never ages out of a contract, and can speak several languages in one afternoon. That's why brands are trying them. It's also why their legal teams are nervous: the rules moved faster than most content teams noticed.

This isn't legal advice. It's how the rules look from inside a studio that makes these films, and what we build into every production so a brand never finds out the hard way.

Two separate things, and brands often mix them up. One is about money, the other is about reality.

The paid connection. If a creator or character promotes a brand that paid them, gave them product or any other benefit, the post must be labelled as an ad. ASCI's influencer guidelines list the labels that count (Ad, Advertisement, Sponsored, Collaboration, Partnership and a few more) and say the label can't hide in a pile of hashtags or behind a "more" link.

The synthetic person. ASCI also says a virtual influencer must additionally tell people they aren't interacting with a real human being, upfront and prominently. In the US, the FTC's 2024 rule on fake reviews and testimonials covers testimonials from people who don't exist, AI-generated ones included.

The likeness. Using a real person's face or voice needs their permission, and that doesn't change because AI recreated it. Synthetic doesn't mean rights-free.

For video posts with no caption, ASCI is specific. Under 15 seconds, the label stays for at least 3 seconds. From 15 seconds to 2 minutes, it stays for a third of the video. Over 2 minutes, it stays for the whole section where the brand appears. Both the brand and the creator are responsible.

Beyond influencer posts, realistic synthetic content now carries its own rules. India's February 2026 IT Rules amendment requires prominent labels and provenance metadata on it. The EU AI Act's transparency duties for AI-generated content apply from August 2026. Provenance standards such as C2PA let a file carry a record of how it was made.

You can't check a hundred cutdowns by hand at the end. The disclosure has to be made with the film, not added after a complaint. In practice that means four habits:

Licence first. Every face, voice and likeness is cleared before anything is generated. Label by default. The ad label and the AI disclosure are part of the edit, sized and timed to the rules for each cutdown. A person approves. Nothing ships without a human sign-off. Keep the record. Who approved what, what was disclosed, and when.

There's a creative choice in here too. The audience research we covered earlier found people push back on photoreal synthetic humans pretending to be real, not on good films (audiences punish "looks like AI"). When a human face carries the message, we usually cast a real one, which is also why we run a casting roster of real creators.

Sources: ASCI, Guidelines for Influencer Advertising in Digital Media (clauses 1.2 to 1.5); FTC, final rule banning fake reviews and testimonials (August 2024); EU AI Act, Article 50: transparency obligations; C2PA, content provenance standard; India IT Rules amendment on synthetic content (February 2026).

Related: Wynngrid: AI influencers and UGC, the disclosure playbook