India's AI Labelling Rules 2026: What the IT Rules Require

India now has binding rules on AI-generated audio, images and video. They sit inside the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, which the Ministry of Electronics and Information Technology (MeitY) amended in February 2026.

This guide sets out what the notified text says, who it applies to, and which changes are still only proposals. It is a summary, not legal advice. Check the current text on MeitY's website before you rely on any detail.

Try it: use ProofAI's AI detector for India, free for 10 checks a day with no login.

What is in force, and since when

MeitY notified the amendment on 10 February 2026 as G.S.R. 120(E), and it took effect on 20 February 2026, according to MeitY's consolidated text of the IT Rules and Hogan Lovells. A corrigendum, G.S.R. 148(E), followed on 26 February 2026. A draft of these rules had been released for public consultation in October 2025, as DSCI notes.

The important point: the duties fall on intermediaries, meaning platforms and services that host or spread content. Creators and brands are affected mainly through what platforms now have to ask of them.

What counts as "synthetically generated information"

Rule 2(1)(wa) defines synthetically generated information (SGI) as audio, visual or audio-visual content that is artificially or algorithmically created or altered using a computer resource so that it appears real, and is likely to be seen as indistinguishable from a real person or real event (MeitY consolidated text; AMS Shardul summary).

Want to check your own content? Try ProofAI free 10 checks a day, no login needed.

When done in good faith, these are excluded:

routine or good-faith editing, formatting, enhancement, technical correction, colour adjustment, noise reduction, transcription or compression, as long as it does not materially alter, distort or misrepresent the substance, context or meaning of the original
the use of computer resources solely to improve accessibility, clarity, quality, translation, description, searchability or discoverability, without generating, altering or manipulating any material part of the original
the routine or good-faith creation of documents, presentations, PDF files, educational or training material and research outputs, as long as this does not result in any false document or false electronic record

Note that the definition covers audio and visuals. It does not cover AI-written text on its own.

The main obligations

1. Prominent labels. Under Rule 3(3)(a)(ii), SGI that is allowed on a platform must be prominently labelled. Visual content needs a visible disclosure; audio needs a disclosure played at the start (AMS Shardul).

2. Metadata and provenance. The same rule requires permanent metadata or equivalent provenance markers, including a unique identifier, to the extent technically feasible. Rule 3(3)(b) says platforms must not let users remove or change these labels or markers (MeitY consolidated text).

Also read: AI Detector False Positives: Why Human Writing Gets Flagged

3. Blocking harmful SGI. Rule 3(3)(a)(i) requires reasonable technical measures to prevent prohibited SGI, including child sexual abuse material, non-consensual intimate imagery, false documents and realistic false depictions of people or events.

4. Faster takedowns. Content flagged by a court order or valid government notice must be removed within three hours, down from 36 hours (Rule 3(1)(d)). Complaints about intimate or impersonating images must be acted on within two hours, down from 24 (Rule 3(2)(b)). General grievances must be resolved within seven days instead of 15 (AMS Shardul).

5. Declarations on large platforms. Under Rule 4(1A), significant social media intermediaries must ask users, before publishing, whether content is SGI, verify that answer with technical measures, and label confirmed SGI. A platform that knowingly allows violating SGI is treated as failing its due diligence (MeitY consolidated text). The threshold for "significant" is more than 50 lakh registered users in India (Hogan Lovells).

Platforms that offer AI creation tools must also warn users that misuse can lead to penalties (Rule 3(1)(ca)), and all intermediaries must remind users of their rules at least once every three months instead of once a year (AMS Shardul).

What is still only a proposal

On 30 March 2026, MeitY published a further draft amendment. A MeitY notice dated 21 April 2026 added a proposal to change Rule 3(3)(a)(ii) so that SGI labels are displayed continuously and visibly for the whole duration of the content, and extended comments to 7 May 2026. The draft also proposes making Ministry advisories and directions binding as part of due diligence (MeitY draft; SFLC.in).

When we checked on 16 October 2026, we did not find a notified final version of this draft. Treat the "continuous label" requirement as proposed, not law, until MeitY notifies it.

What this means in practice

Creators and influencers - Answer platform AI-disclosure prompts honestly. Large platforms are now required to ask and to verify. - Expect labels on realistic AI faces, voices and scenes, even in parody.

Brands and agencies - Ask agencies and freelancers to declare any AI-generated visuals, voices or spokesperson videos in deliverables. - Keep the original files your AI tools export. Re-saving or screenshotting can strip metadata.

Platforms and apps - Map which features create or host SGI. Build labelling, metadata preservation and, if you are an SSMI, declaration and verification flows. - Set up an on-call process that can meet three-hour and two-hour deadlines.

Newsrooms and publishers - Label AI illustrations clearly. - Verify user-submitted images and audio before publishing, especially during elections or emergencies.

Where AI detection fits

The rules rely on labels and metadata, but both depend on the creating tool adding them, and metadata can be lost along the way. Content-based detection is a second check for files that arrive without any provenance information. It gives a likelihood, not a legal finding, so pair it with source checks and human review.

If you need to screen images before posting or approving them, ProofAI's AI image detector accepts JPG, PNG and WebP files up to 25 MB, with 10 free checks a day and no account needed. ProofAI results are probabilities, not proof, so use them as one signal in your review.