India's binding rules for synthetic content — a transparency regime without an AI law

Six months ago India made a deliberate choice not to pass a horizontal AI statute, betting instead on existing law, guidance and new institutions — the posture set out in its AI Governance Guidelines. That choice has now produced its first hard edge. On 10 February 2026 the Ministry of Electronics and Information Technology (MeitY) notified G.S.R. 120(E), the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, which took effect on 20 February 2026. The instrument is not an AI act. It is an amendment to the 2021 intermediary rules, made under section 87 of the Information Technology Act, 2000 — and through it India has built one of the world's first binding transparency-and-provenance regimes for AI-generated content.

Defining the object: 'synthetically generated information'

The rules add a new definition in rule 2(1). 'Synthetically generated information' (SGI) means "audio, visual or audio-visual information which is artificially or algorithmically created, generated, modified or altered using a computer resource, in a manner that such information appears to be real, authentic or true and depicts or portrays any individual or event in a manner that is, or is likely to be perceived as indistinguishable from a natural person or real-world event." A companion definition covers "audio, visual or audio-visual information" itself.

The scope is bounded by carve-outs. Routine, good-faith editing — colour adjustment, noise reduction, transcription, compression — does not make content SGI, provided it does not "materially alter, distort, or misrepresent the substance, context, or meaning" of the original. Nor does the ordinary creation of documents, PDFs, presentations, educational material or research outputs, unless it produces a false document or false electronic record; nor do purely accessibility-oriented adaptations. The definition therefore targets deceptive synthetic media, not the everyday use of generative tools.

Two duties: provenance at creation, verification at distribution

The substance sits in two new obligations, both engineered to operate through the intermediary safe-harbour conditionality of section 79 of the IT Act rather than as free-standing AI duties.

Creation tools (new rule 3(3)). An intermediary that "offers a computer resource which may enable, permit, or facilitate" the creation of SGI must do two things. First, deploy "reasonable and appropriate technical measures, including automated tools," to stop users generating unlawful SGI — content that is child sexual abuse material or non-consensual intimate imagery, that creates a false document, that concerns explosives or arms, or that "falsely depicts or portrays a natural person or real-world event" in a manner likely to deceive. Second, every other piece of SGI must be "prominently labelled" — visibly in the display, or through a prefixed audio disclosure — and "embedded with a permanent metadata or other appropriate technical provenance mechanisms, to the extent technically feasible, including a unique identifier" tracing the computer resource used. Crucially, the intermediary "shall not enable the modification, suppression or removal" of that label or metadata.

Significant social media intermediaries (new rule 4(1A)). Before any user content is displayed, uploaded or published, an SSMI must (a) require the user to declare whether the information is synthetically generated, (b) "deploy appropriate technical measures … to verify the accuracy of such declaration," and (c) where the content is confirmed as SGI, ensure it carries a clear, prominent label. An explanation makes the standard "reasonable and proportionate technical measures" — verification, not strict liability, but an affirmative duty nonetheless.

Notably, the final rules contain no fixed-percentage labelling rule. Earlier drafts and much secondary commentary referred to a label covering "10% of the display surface" or the opening tenth of audio; that quantitative test was dropped in favour of the qualitative "prominent visibility" standard above.

Sharper enforcement timelines

The amendment also compresses the removal windows that give the obligations teeth. Orders to disable access under rule 3(1)(d) move from thirty-six hours to within three hours; the general grievance-removal window in rule 3(2)(a) falls from fifteen days to seven days (with its inner proviso cut from seventy-two to thirty-six hours); and the window for specified categories in rule 3(2)(b) drops from twenty-four hours to two hours. Rule 7's reference to the Indian Penal Code is replaced by the Bharatiya Nyaya Sanhita, 2023.

A third model for the synthetic-media problem

The regulatory goal — letting people know when content is machine-made — is now shared across the major jurisdictions, but the instruments diverge. The EU reaches it through Article 50 of the AI Act, which from 2 August 2026 requires providers to mark synthetic output in a machine-readable format and deployers to disclose deepfakes. China reached it through dedicated content-labelling measures in force since September 2025. India has reached it through a third route entirely: not an AI law, but the liability architecture that already governs its platforms, hard-wiring labelling, provenance and verification into the conditions for safe harbour.

For organisations operating across borders, the lesson echoes the one running through our Asia-Pacific comparison: the destinations converge on transparency, but the legal vehicles do not — and a provider that satisfies the EU's machine-readable-marking duty has not thereby satisfied India's prominent-label-plus-embedded-provenance duty, still less its two-hour and three-hour takedown clocks.