Notes from the lab · What's legal

India's AI disclosure rules, and what a brand actually has to do

July 2026 · 7 min read · Wynngrid

Every Influencer OS conversation reaches legal within two meetings. Fair enough. AI faces in advertising sit under three separate sets of Indian rules, and most summaries mix them up. Here they are, separated, with the part a brand actually has to do.

Layer 1: the IT Rules and the synthetic content amendments

The IT Rules 2021 put obligations on platforms and intermediaries around synthetic media, and MeitY has been tightening that screw since the 2023 deepfake advisories. The material development for brands: the draft amendment circulated in late 2025 proposes that synthetically generated content carry a visible label, with the draft suggesting the label cover a meaningful share of the visual, about a tenth of the frame in the proposal. The final shape is still pending as we write this.

What this means practically: the compliance burden lands first on platforms, which means platforms will push it onto advertisers through their upload tools. Instagram and YouTube already ask whether content is AI-generated. Answer honestly. A brand that labels today is compliant with both the current advisories and the likely final rule.

Layer 2: ASCI

The Advertising Standards Council of India cares about two things here. First, the general rule that an ad must be identifiable as an ad, which is why influencer posts carry material-connection labels like #ad. Second, the specific rule for virtual influencers: consumers must be told they are not interacting with a real human being. The disclosure has to be upfront and prominent, not buried in a bio.

ASCI is self-regulatory, but treat it as binding. Its complaints process is public, fast, and picked up by the press, and CCPA has shown it will follow where ASCI points.

Layer 3: personality rights

This is the layer that actually gets companies sued. Indian courts, Delhi High Court in particular, have repeatedly protected celebrities against AI use of their face and voice: the Anil Kapoor and Amitabh Bachchan orders are the reference points. If a generated face or voice resembles a real person and you have no license, you carry that risk with every impression.

The safe positions are two: a licensed likeness with a contract covering AI generation specifically, or a fully invented face with documentation proving it was invented. Resemblance you cannot explain is a liability you cannot price.

What a brand actually has to do

1. Label AI-generated assets, both in the creative where prominence demands it and in the platform's own AI-disclosure toggles.
2. Disclose virtual influencers as not human, prominently, on the ad itself.
3. Keep provenance records: what was generated, with which model, from which brief, approved by whom. If a regulator or platform asks, you answer in a day, not a month.
4. License every face and voice, with AI-specific language in the contract. For invented faces, keep the generation trail.
5. Keep the standard ASCI influencer disclosures on paid posts. AI does not exempt you from #ad.
6. Re-check quarterly. The late-2025 draft will land in some form, and label formats may change.

How we run it

Every asset that leaves Influencer OS ships disclosed under the IT Rules advisories and ASCI guidelines by default, with provenance logged at generation time. Compliance is a pipeline step, not a review meeting. That is the whole trick: make the legal checklist something the system does, and legal stops being the department that says no.