India’s effort to control deepfakes has moved beyond advisories. The updated Information Technology Rules now define a class of synthetically generated information and attach practical duties to the platforms that distribute it and the services that help users create it.
The change matters because the compliance question is no longer simply whether a piece of AI content is unlawful. Platforms and creation tools must also think about disclosure, traceability and how users are warned before realistic synthetic media is published.
What counts as synthetically generated information?
MeitY’s official FAQ says the definition is focused on artificially created or altered audio, visual or audio-visual material that appears real or authentic and may be perceived as indistinguishable from a natural person or real-world event. Pure text, by itself, is not covered by this specific definition.
That distinction is important. A routine AI-written caption and a realistic cloned-voice video do not create the same disclosure risk. The rules target the type of synthetic media most capable of misleading a viewer about who spoke, what happened or whether an event was real.
What platforms are expected to do
The consolidated rules and MeitY guidance establish a stronger due-diligence framework. Relevant intermediaries need processes that can identify covered synthetic media, preserve required provenance information and display a prominent label. Platforms that enable creation or modification of synthetic media also face additional user-awareness and verification duties.
For product teams, this turns compliance into a workflow problem. A label added only after a complaint is unlikely to be enough. The disclosure has to survive upload, compression, reposting and display across different screen sizes. Metadata must be treated as operational data, not decorative information that disappears during transcoding.
What creators, agencies and brands should change
Creators may not carry every obligation imposed on an intermediary, but their publishing process now needs a reliable disclosure step. Agencies should keep the original generation file, prompt record, editing history, consent documentation and final labelled export together. Brands using a cloned voice or synthetic spokesperson should document permission and make the disclosure visible before publication.
- Classify AI assets as text, image, audio or video before publishing.
- Flag realistic impersonation or event simulation for human review.
- Add a clear viewer-facing label; do not rely only on a caption or hashtag.
- Retain provenance metadata and a clean audit trail.
- Record consent when a real person’s face, likeness or voice is used.
- Prepare a correction and takedown process for disputed media.
The practical business consequence
The biggest impact may be on the design of AI creation products. Tools serving Indian users will increasingly need disclosure controls at the point of export, not as an optional setting buried in a menu. Social platforms will need consistent treatment across uploads and reshares, while advertisers will need contractual assurances from production partners.
This also creates an opportunity. Platforms that make provenance easy can reduce moderation costs and give responsible creators a way to demonstrate authenticity. In a market flooded with low-cost AI media, a trustworthy content chain can become a product feature.
The central shift is from asking whether AI was used to asking whether realistic synthetic media can be identified, explained and traced.
What remains uncertain
Edge cases will still require judgement. Stylised animation, obvious parody, minor retouching and mixed human-AI production may not present the same risk as a photorealistic impersonation. Enforcement practice will determine how consistently platforms interpret the threshold.
Businesses should therefore avoid treating this article as legal advice. The safer operational approach is to review the official notification and consolidated rules, obtain counsel for high-risk deployments and build a disclosure system that can be updated without redesigning the whole product.
Sources and further reading
- MeitY FAQ on the 2026 amendments — definition, scope, commencement and practical clarifications.
- Consolidated IT Rules updated 10 February 2026 — controlling regulatory text.
