When does India switch off the internet, and does it actually work?

What apps has India actually switched off?

In June 2026, the Ministry of Electronics and Information Technology (MeitYMeitY (Ministry of Electronics and Information Technology)The Union ministry that oversees India's digital policy, issues blocking orders under Section 69A, and sometimes discloses numbers to Parliament. Its figures are self-reported and occasionally inconsistent.All URL blocking numbers come from MeitY's own disclosures; the article notes these are self-reported and sometimes inconsistent across different answers.) temporarily restricted access to Telegram across India under Section 69A of the IT Act, acting on a request from the National Testing Agency. The reason: fears that leaked NEET-UG re-exam papers were circulating on the platform. The restriction, set to lift on 22 June once the re-exam was over, reached a service Telegram says has about 150 million Indian users. Telegram petitioned the Delhi High Court on 17 June 2026, arguing that Section 69A allows blocking specific content, not switching off an entire messaging service. On 19 June the Court disagreed: it upheld the block as a proportionate, temporary measure and held that Section 69A does empower the government to switch off a whole platform. It was the first serious judicial test of that power, and the power won.

But Telegram was not the first app to vanish. India has a standing power to block whole apps, and it has used it in waves. In June 2020, days after the Galwan clash, 59 Chinese apps including TikTok were blocked in one action. Another 47 clone apps followed in July, and 118 apps, including PUBG, in September. In February 2022, 54 more Chinese apps were blocked. In February 2023, about 232 betting and loan apps, many Chinese-linked, were taken down in one of the largest single actions. In April 2023, 14 messenger apps were blocked in Jammu and Kashmir over alleged terror use. Then in June 2026, just one app, Telegram, became the latest and most consequential block because of the legal challenge it sparked.

These are not local connectivity kills. They are whole-app blocks under Section 69A, leaving the internet itself on but making a specific platform unreachable. This is one of three distinct off-switches India uses. The other two are shutdowns, cutting connectivity itself in a place, and DNS blockingDNS BLOCKINGHow internet service providers usually carry out a blocking order: when you type a web address, your ISP looks it up in a directory; if the government has told the ISP to block that address, the ISP simply doesn't give back the right number, and the site never loads. It is like a phone operator refusing to connect a call. DNS blocking is easily bypassed by using a different DNS server or a VPN.This is why the same blocked site may be reachable on one ISP but not another, and why the blocks in the dnsblocks.in study reflect intent rather than airtight censorship., the technical means ISPs use to implement a block. Understanding the difference matters because the tools, and their consequences, are not the same.

How often does India actually pull the internet plug?

Since 2012, India has ordered about 922 government-mandated internet shutdowns, as documented by the SFLC.inSFLC.in (Software Freedom Law Centre, India)A legal-technology non-profit that, among other things, maintains a public tracker of every reported internet shutdown in India, based on news reports, court filings, and government orders. It is the closest thing to an official tally, since the government does not publish a complete list.SFLC's tracker is the primary source for the shutdown counts used throughout the article, and understanding it as a non-official, minimum count is essential. Internet Shutdowns Tracker. In 2012, there were just 3 shutdowns. That number climbed steeply, peaking at 136 in 2018, a year when rumour-driven lynchings and protests pushed states to pull the plug repeatedly. After that, the annual count began to fall: court scrutiny tightened, especially after the Supreme Court’s Anuradha Bhasin ruling in 2020 required shutdowns to be temporary, necessary and proportionate, and by 2024 the recorded count had dropped to 60, and to 54 in 2025. In 2026, as of the data cut-off, 20 shutdowns have been recorded, but this is a partial year-to-date figure and will rise by December.

These numbers are not the full picture. Every count is a documented floor because the government rarely publishes shutdown orders proactively. The SFLC.in tracker relies on news reports, court filings, and occasional official confirmation; shutdowns that go unreported in local media are simply not counted. So the true number of times some part of India has gone dark by order is almost certainly higher. The decline since 2018 is real, but fewer shutdowns can still mean longer ones, a single blackout lasting weeks does more damage than a dozen brief ones, and the data on duration, as we will see, remains patchy.

Where does India go dark?

The national count of 922 shutdowns hides a stark geography: shutdowns are not spread across India, they are concentrated in a few places. A choropleth map of cumulative shutdowns from 2012 to 2026 shows Jammu and Kashmir painted in the darkest shade, with some other states like Rajasthan and Manipur also standing out, while most of India has almost none. The all-India figure is largely driven by one region in repeated crisis; it is not a national habit but a localised emergency used in conflict-prone areas. The map reframes the problem instantly: the internet does not go dark everywhere equally, and the places that suffer most are often those already under strain.

Which states have lost the internet the most?

When we list them, the ranking is stark. Jammu and Kashmir’s 453 shutdowns over 15 years mean the region has spent a substantial portion of that period offline by government order. The reasons are rooted in its long insurgency, political sensitivities, and multiple flashpoints, elections, anniversaries, protests, that prompt authorities to reach for the off switch. Rajasthan’s 114 shutdowns, the second highest, are largely tied to competitive exams and protests; the state has made a habit of cutting the internet to prevent cheating, sometimes affecting entire districts.

Manipur’s 62 shutdowns reflect ethnic tensions and prolonged conflicts, often involving extended outages. Haryana (47), Uttar Pradesh (37), and West Bengal (27) show that even politically volatile states in the heartland use the tool, though far less than the conflict zone. Bihar and Odisha each have 24, Maharashtra 19, and Meghalaya 18. The tail is long, but the numbers drop sharply after the top five. This league table makes plain that the internet off switch is not a nationwide emergency tool applied evenly; it is a routine administrative reflex in a handful of states, and an extraordinary measure elsewhere.

Are these blackouts mostly a reaction to violence, or a reflex before anything happens?

This is where the official justification meets the data head-on. By SFLC.in’s classification, most Indian shutdowns are preventive, imposed in anticipation of an event, not in reaction to unfolding violence. In 2021, the split was 94 preventive shutdowns to just 6 reactive. By 2024, it narrowed to about 33 preventive versus 27 reactive, but the overall pattern remains: the state pulls the plug before trouble starts, not during a crisis. The reasons are often stated: a tense exam day, a planned protest, a sensitive anniversary. But the official defence of shutdowns rests on public safety in an emergency, and a mostly pre-emptive pattern sits awkwardly with that.

Does this pre-emption work? A 2019 study by Jan Rydzak, ‘Of Blackouts and Bandhs’, is the most careful empirical work on Indian shutdowns. It found that shutdowns were associated with more violent collective action, not less. The mechanism is intuitive: cutting communication does not stop people mobilising; it pushes coordination from non-violent, organized tactics toward decentralized, spontaneous, and often violent ones. So the justification governments lean on most, that shutdowns prevent violence, is the one the best evidence most directly contradicts.

But let us be fair: the problems are real. In 2018, a wave of mob lynchings killed more than a dozen people, fuelled by false child-kidnapping rumours on WhatsApp. Police cut connectivity in places to stop a rumour that could get someone killed within hours. Exam-paper leaks via Telegram and similar channels are a documented, recurring problem, exactly the fear invoked in the 2026 NEET block. The honest framing is not that shutdowns are evil, but that these are real dangers met with a sledgehammer: blunt, collective punishment of millions for the acts of a few, with evidence it can backfire.

How long do these blackouts last, and what does the state actually tell us about them?

Even basic facts about a shutdown are often hidden. Across the 922 recorded shutdowns, the single largest duration bucket is ‘never disclosed’: 210 shutdowns where nobody outside government knows how long the internet stayed off. Another 353 lasted under 24 hours, 245 between 24 and 72 hours, and 114 stretched beyond three days. But the opacity around the 210 is a finding in itself. Orders are rarely published, and the length of a blackout can only be guessed from news reports that seldom say when the internet returned.

The Supreme Court’s Anuradha Bhasin judgment (January 2020) was supposed to change this. The Court held that the freedom to use the internet for speech and for trade is protected under Article 19 of the Constitution, that any suspension must be temporary, necessary, and proportionate, and that the government must publish the orders. Years later, that transparency requirement remains largely unmet. Independent trackers like SFLC.in are the public’s only real record, and even they cannot fill the gap when the state stays silent. Without knowing how long a shutdown lasted, the public cannot evaluate whether it was proportionate, the very test the Court set. Opacity is not just a technical flaw; it is the condition that makes a tool unaccountable.

What about the quieter off switch, how many websites and accounts does India block?

Shutdowns are visible and noisy; URL blocking under Section 69A is silent and far larger. As disclosed to Parliament by MeitY, the number of websites, accounts, and URLs ordered blocked grew from 1,385 in 2017 to a peak of 9,849 in 2020. It has stayed in the thousands every year since, with 7,502 blocks in 2023. These are the government’s own figures, and even they do not fully reconcile across different parliamentary answers, 2019 has been stated as both 3,635 and 3,655. The orders themselves are confidential under Rule 16 of the 2009 Blocking Rules, so the public never learns what was blocked or why.

This machinery dwarfs shutdowns in scale but operates invisibly. A citizen in Delhi may never know that a website just became unreachable, because there is no public notification. But even confidential totals can be prised apart two ways: by who issues the orders, and by which platforms they land on.

So who is actually ordering all this blocking?

Picture website blocking and you probably picture a judge. In reality the executive does roughly half of it, in secret. SFLC.in's 'Finding 404' report, pieced together from right-to-information replies, counted about 55,580 websites blocked in India between January 2015 and September 2022. Court orders, overwhelmingly for copyright piracy, account for 46.8 percent of those. Executive orders under Section 69A account for 47.5 percent, almost exactly the same share. The two routes run neck-and-neck, but they differ in one way that matters: court orders are public and reasoned, while Section 69A orders are confidential by rule. You can read why a court blocked a pirate-streaming site; you cannot read why the government blocked anything.

That gap matters because the two halves do different work. The court-ordered pile is mostly anti-piracy, and a handful of sweeping orders each take down hundreds of mirror sites, which puffs up the count. The Section 69A pile is where blocking on grounds of sovereignty, security and public order lives, the blocking that actually touches speech, and it is the half nobody outside government gets to inspect. An evenly split chart, then, understates the worry: the opaque half is the one that can reach a news site or a critic.

And which platforms get blocked the most?

The aggregate count says nothing about who gets censored. A Rajya Sabha answer in December 2023 (Unstarred Question 732) finally broke the Section 69A numbers down by platform, and the picture is a clean crossover. In 2018 and 2019, Facebook drew the most takedowns by far: 1,555 URLs in 2018 against just 224 for what was then Twitter. From 2020 that flipped hard. Blocks aimed at X (Twitter) jumped to 2,731 in 2020 and kept climbing, peaking at 3,423 in 2022, while Facebook stayed high but fell behind. YouTube spiked in the 2020 pandemic year and then settled.

The flip is not really about platform size; it tracks the politics. X's surge lines up with the most contested stretch of recent Indian speech online: the farmers' protests of 2020-21, the government's public standoff with Twitter over blocking accounts, and the running fights over political content. The platform that became the venue for dissent became the platform the state blocked most. These are counts of URLs, accounts and posts rather than unique pieces of content, and the 2023 figure runs only to October, but the direction is unmistakable.

What is actually being blocked when an Indian ISP refuses to connect to a domain?

The answer complicates the easy story. The dnsblocks.in ‘Poisoned Wells’ study examined 43,083 domains found DNS-blocked across six major Indian ISPs. The overwhelming majority are piracy and streaming: Movies and TV alone accounts for 20,986 domains, file sharing 2,188, and live-streaming piracy 1,224. Pornography blocks number 2,953, and gambling 1,906. Malware and IP infringement add 921 and 521 domains respectively.

What does not dominate is political censorship. Content tied to speech and access is a tiny sliver: news media around 30 domains, political criticism around 10, government around 8, circumvention tools around 8. This does not mean the government never targets speech, DNS blocking is just one tool, and there is a large ‘uncategorised’ bucket of 10,027 domains that remains unclassified. But the weight of DNS blocks is overwhelmingly copyright and vice enforcement, not political silencing. That said, DNS blocking is trivially bypassed: anyone with a VPN or an alternative DNS server can reach these domains, so the measure reflects intent more than an effective wall.

What does switching off the internet actually cost India?

The most authoritative India-specific estimate comes from the ICRIERICRIER (Indian Council for Research on International Economic Relations)An economic policy think tank that produced the most cited study on the economic cost of Indian internet shutdowns, 'The Anatomy of an Internet Blackout' (2018). It used an input-output model to estimate lost output.ICRIER's $3 billion estimate over six years is the anchor for the cost discussion; understanding it is a model, not a receipt, is crucial. working paper ‘The Anatomy of an Internet Blackout’ (2018), led by Kathuria, Kedia and others. It examined about 16,315 hours of shutdowns over 2012 to 2017 and put the economic loss at roughly 3 billion dollars. Mobile-only shutdowns accounted for about $2.37 billion of that, and combined mobile-and-fixed shutdowns roughly $680 million more. That is about four-fifths from mobile cuts alone, reflecting that most Indian shutdowns target mobile data.

These are macroeconometric model estimates, not measured losses from company balance sheets. They work by assigning a GDP loss rate to each hour offline. For recent years, the NetBlocks and Internet Society Cost of Shutdown Tool, built on the Brookings method, puts annual losses in the hundreds of millions of dollars. But beyond the macro numbers lies a human cost that hits the poorest hardest. India spent a decade pushing its informal economy onto digital rails, UPI payments, mobile orders, online welfare. A shutdown then strands exactly those people: the small trader who can no longer take a payment, the gig worker who cannot get a ride, the patient whose hospital cannot process an Ayushman Bharat claim. During the 2019-20 Kashmir blackout, local trade bodies estimated the loss at well over a billion dollars, idling tens of thousands of artisans. The distributional sting is that the state’s own digital inclusion drive makes the shutdown’s harm more severe for those it was meant to connect.

Where these numbers come from, and what they can't show

Every count on this page is a documented floor. Shutdown data comes from the SFLC.in Internet Shutdowns Tracker, which collates reports from news, court filings, and official orders. DNS blocking data is from the dnsblocks.in ‘Poisoned Wells’ study, which probed six ISPs. App ban counts are from MeitY notifications and press reporting. URL blocking numbers are from MeitY's own answers to Parliament, including the platform-wise breakdown in Rajya Sabha Unstarred Question 732 (8 December 2023); the split between executive and court-ordered website blocks is from SFLC.in's 'Finding 404' report, built on right-to-information replies. The economic cost estimates are from the ICRIER study and, for recent years, the Cost of Shutdown Tool. The legal benchmark is the Supreme Court’s Anuradha Bhasin judgment.

Because shutdown and blocking orders are secret, the true totals are higher than what you see here. Independent trackers also legitimately disagree: Access Now counted 84 Indian shutdowns in 2024, while SFLC recorded about 60, largely because they define a shutdown differently. That gap is information, not error, it tells us how hard this phenomenon is to measure. The cost figures are model estimates, not measured losses. And the June 2026 Telegram block has now been tested: on 19 June the Delhi High Court upheld it and confirmed that Section 69A reaches whole platforms, not just individual posts, giving the executive's broadest off switch a court's blessing. The data gives us a floor, not a ceiling, and the picture is sharp enough to act on, but never complete.