Meta restricted more than 5.4 million posts and accounts in India under local law in the first half of 2026, up from roughly 70,000 in the previous six months, according to Reuters reporting on the company's transparency disclosures, as summarised by The South First. That is roughly a 77-fold increase, and it puts India second only to Indonesia among countries where Meta restricts content under local law. Meta said most of the actions were cases where it was "obligated to proactively restrict content based on local law requirements."
The number is not a measure of how much unlawful content India has. It measures what happens when the legal cost of delay rises sharply.
The strongest case for the government
The government's argument deserves a fair statement. Deepfakes, non-consensual intimate imagery and incitement can do irreversible harm within hours. A 36-hour window, the earlier standard, is a long time for a viral fake. The amendment to the IT Rules was notified on 10 February 2026 and took effect on 20 February, according to Bar & Bench. It also introduced a definition of synthetically generated information and labelling duties. New Delhi says it orders takedowns only where content violates existing law or threatens public safety or national security. Platforms with large resources can plausibly meet a tight deadline for clear-cut harms. That reasoning is serious and should not be waved away.
What actually changed
Bar & Bench reports that the amendment cut the deadline for complying with lawful takedown directions from 36 hours to three. It also shortened grievance disposal from 15 days to 7. SCC Online describes the obligation as removing or disabling access within three hours of receiving actual knowledge, covering content tied to unlawful acts such as defamation or contempt of court, or information prohibited under any law. The rules are catalogued by the Digital Policy Alert as G.S.R. 120(E).
The South First, citing DW, adds that authorised police officials can notify intermediaries of content they consider unlawful without a formal complaint or other legal proceedings, leaving companies very little time to review. I could not confirm the exact rank requirements from the gazette text, so this analysis does not rely on them.
Why the spike is predictable
Safe harbour under Section 79 of the IT Act is the core of the intermediary bargain: a platform is not liable for user content if it acts properly once it has knowledge. When the window to act shrinks to three hours and the notice can come from a police official rather than a court, the rational platform response is to remove first and review later, or never.
That is what Meta's own wording suggests. Restrictions where it was "obligated to proactively restrict" are a compliance posture, not an adjudication that each post was illegal. A review team cannot assess legality, context, satire or public interest in three hours across millions of items. Automated matching and blanket compliance fill the gap.
The reported content mix reinforces the concern. The South First reports that restricted material allegedly included criticism of the government, the BJP and the RSS from opposition figures, journalists and creators, and even cartoons of the prime minister. Those details are reported claims, not findings, and the government disputes any political motive. But the structure of the rule makes over-removal likely whatever the intent.
The constitutional baseline
India has already litigated this question. In Shreya Singhal v. Union of India (2015), the Supreme Court struck down Section 66A as vague and unconstitutional under Article 19(1)(a). It then read down Section 79(3)(b) and Rule 3(4) of the 2011 intermediary rules. Under that reading, an intermediary loses its exemption only if it fails to act after receiving actual knowledge of a court order, or a government notification limited to unlawful acts relatable to Article 19(2). The Court reasoned that platforms cannot be asked to judge the legality of millions of requests themselves.
A three-hour clock combined with direct police notices sits uneasily with that logic. The Court was concerned with exactly this pressure to over-comply, and it tied removal duties to a formal, reviewable trigger. Whether the new rules honour that line is a question courts will have to answer. The numbers now give them an empirical record to work from.
A proportionate alternative
The goal of fast removal for harmful content is defensible. The design can be fixed without abandoning it.
- Tier the clock by harm. Short windows suit narrow, verifiable categories such as non-consensual intimate imagery and clearly identified deepfakes. Political, satirical or commentary content should keep longer review periods.
- Require reasoned, reviewable orders. Every direct notice should state the legal basis and the Article 19(2) ground, and be logged for audit.
- Notify affected users. Account holders should be told what was removed and why, with a route to appeal. Meta's restoration of an opposition party account, reported by The South First, shows errors do happen.
- Publish government-side data. Platforms report their numbers. The authorities issuing the notices should report theirs, including how many were later withdrawn.
The takeaway
A jump from about 70,000 to over 5.4 million restrictions in one half-year is a regulatory outcome, not an organic shift in online behaviour. Fast takedown has a legitimate place in a modern intermediary regime. When the timeline is measured in hours and the trigger is a police notice, the costs fall on lawful speech. A proportionate regime would target the narrow harms that justify speed and leave room for the review that Shreya Singhal treated as essential.