India’s AI video labelling rules, explained for businesses

Since 20 February 2026, realistic AI-generated video published in India must carry a visible AI label and embedded provenance metadata. Here’s what that looks like in practice, without the legalese.

The core obligation

MeitY’s February 2026 amendment to the IT Rules treats realistic AI-generated media as “synthetically generated information.” Anyone publishing it must label it clearly and attach provenance metadata that records where it came from. The intent is simple: a viewer should be able to tell it’s AI, and a platform should be able to trace it.

What a compliant label looks like

A visible disclosure on the video that it is AI-generated. The current draft under consultation would require that label to remain continuously and clearly visible throughout the video’s duration — so a one-frame watermark at the start is unlikely to be enough going forward.

What provenance metadata means

Machine-readable data embedded in the file itself that identifies the content as AI-generated and records its origin, so the label can’t simply be cropped away without a trace. Platforms are expected to preserve, not strip, this metadata.

What to keep on file

Keep the labelled master, the provenance metadata, and — if a real person’s face or voice is used — a record of their consent. MirrorProd issues all three as a Certificate of Provenance with every delivery, so the paper trail exists before you need it.

Rules status as of 13 Jul 2026 — labelling and provenance obligations effective 20 Feb 2026; final label-visibility norms still in consultation. This is a delivery standard, not legal advice.