Constant AI labelling, lasting pain: Why govt's label mandate could reshape social media

On April 21, MeitY added what may be the most operationally demanding compliance requirement yet proposed for social media (SM) platforms in India.

Through a late-stage insertion into draft amendments to IT (Intermediary Guidelines and Digital Media Ethics Code) Rules 2021, the ministry proposed that intermediaries must ensure 'continuous and clearly visible display of such label(s) throughout the duration of the content, in a visual display' for synthetically generated information (SGI). The public comment deadline was extended from April 29 to May 7.

A watermark in the opening seconds of a video or a disclosure buried in a caption would no longer suffice. The label must stay on-screen for the entirety of playback. In a country with over 750 mn internet users and hundreds of millions of short-form video consumers, this is not a marginal adjustment. It is an architectural demand.


This proposal sits atop a rapidly layered regulatory edifice. On February 10, MeitY notified IT Amendment Rules 2026, introducing SGI - audio, visual or audio-visual content artificially created to appear indistinguishable from a natural person or real-world event.

Those rules required SGI to be 'prominently labelled' with permanent metadata and cut takedown timeline for unlawful synthetic content from 36 hrs to 3 hrs. They are a response to a draft from October 2025 that civil society criticised for overbreadth and over-reliance on AI detection tools described as 'empirically proven to work poorly'. MeitY narrowed the definition in February, carving out routine editing and good-faith creation. The April 21 addition moves goalposts back toward greater stringency.

A persistent on-screen label sounds like a minor UX tweak. It is not. The mandate spans 15-sec Instagram reels, hour-long AI-assisted YouTube documentaries, AI-generated regional-language news clips, and AI-filtered WhatsApp images.

For significant SM intermediaries (platforms with over 5 mn Indian users), the burden is already heavy - user declarations before synthetic content goes live, technical verification and tamper-proof labels. Continuous visibility means the label must survive every re-share, re-encoding and cross-device viewing instance.

The technical demands go deeper. Metadata-based labelling - an industry standard - deployed through Coalition for Content Provenance and Authenticity framework by platforms like YouTube and Meta is machine-readable but invisible to human viewers.

MeitY's proposal requires a parallel, always-visible rendering-layer overlay. Platforms must maintain both systems in sync. Meanwhile, the 3-hr takedown window for unlawful SGI compels investment in automated detection, tools that remain unreliable for lightly edited or compressed content.

The framework punishes both errors: over-labelling raises compelled speech concerns, and under-labelling strips platforms of safe harbour immunity under Section 79 of IT Act.

India is not alone, but it is moving the furthest. The EU's Article 50 of the AI Act, enforceable from August, requires machine-readable marking by AI providers and human-visible disclosure by deployers of deepfakes in public-interest contexts, with only minimal disclosure required for artistic or satirical content.

Its voluntary Code of Practice is being developed through a 7-mth stakeholder process. Crucially, the EU does not mandate continuous on-screen labels throughout the video duration. China's rules, among the world's most prescriptive, require visible watermarks and disclosure at the start of content, but stop there.

India's proposal to require labelling for the full duration is, in this specific respect, without an international parallel. The US relies on California's SB 942, which requires labels that are 'extraordinarily difficult to remove', and on voluntary systems from Meta, YouTube and TikTok.

MeitY's framework derives its force from one instrument: safe harbour conditionality. Separately, amendments proposed on March 30 would make compliance with any ministry-issued advisory or guideline a due diligence obligation under Section 79. Non-compliant platforms risk losing their liability shield entirely, thus concentrating content governance in the executive branch without proportionate judicial oversight.

A continuous-label mandate, expensive to build and difficult to verify, falls hardest on smaller platforms, cementing structural advantages for large incumbents.

Regulatory asymmetry is rarely neutral. MeitY deserves credit for extending the comment window. What stakeholders do with it will matter enormously. Clearly, they must help MeitY address the worrying issue around deepfakes in a more prudent manner.