India Gives Meta Three Hours

India's IT ministry has issued Meta a formal warning over AI-generated content. The specifics matter more than the headline, because this is not a general appeal for platforms to do better. It is a compliance demand with a clock on it.

What set it off

Meta temporarily removed a video posted by Prime Minister Narendra Modi. That prompted a government summons and a meeting in New Delhi between senior officials on both sides.

MeitY Secretary S. Krishnan delivered what was reported as a clear warning that compliance cannot be optional. Should platforms fail to act on unlawful AI content, bot-driven misinformation and CSAM-related lapses, the government has said it will invoke blocking powers, court directions and police action.

Three hours

The operative number sits in the Intermediary Guidelines and Digital Media Ethics Code Amendment Rules, 2026: a three-hour takedown window. Miss it and a platform risks losing safe harbour, which brings potential criminal liability with it.

Three hours is short. Short enough that no human review process can meet it reliably at national scale. So the rule is, in effect, a requirement to automate moderation and live with whatever error rate that produces.

Few people in this debate want to say that part out loud. A window this tight does not simply demand faster moderation. It demands moderation decisions made by systems that will be wrong in both directions — leaving up what should come down, taking down what should stay up.

The three areas MeitY named

AI misinformation at scale. Not individual bad posts. The volume that synthetic generation makes possible.

Design features that enable crime. A pointed choice of words, moving the argument from content moderation into product architecture.

Failure of safe harbour, tied to the three-hour rule.

The ministry also asked Meta to retool its algorithms and hand over an actual roadmap to compliance rather than a statement of intent.

The legal instrument

Section 69A of the IT Act 2000 lets the central government block public access to online information to protect sovereignty, security and public order. It reaches past individual pieces of content — entire platforms sit within its scope.

Earlier this year the government amended the intermediary rules to bring synthetically generated information under a specific regulatory framework, adding due-diligence obligations for intermediaries.

Reading this from the enterprise side

If you run AI systems touching Indian users, the direction here is legible enough to plan against. The regulatory question is moving from can you detect synthetic content to can you act on it inside a fixed window, and can you show your work afterwards.

That is an engineering and audit-trail problem wearing policy clothes. It implies logging, provenance tracking and demonstrable response times, all of which are far cheaper to design in than to retrofit once somebody is asking for evidence.

Anyone deploying generative AI in India should treat this as a preview rather than a story about someone else's compliance trouble.

Source: India warns Meta to curb AI-generated content or face enforcement