Kill Switches: Internet Shutdowns as a Tool of Governance

A tool no printing press or radio jammer ever had

Before the internet, every regime that wanted to control information had to work at the level of the message. It could license which books a printer was allowed to set, jam a single broadcast frequency, seize a press, or station a censor at a border crossing to inspect what came through. Each of those tools acted on one channel or one piece of content at a time. None of them could reach into a household's phone line, a hospital's dispatch system, and a bank's payment network at the same moment and cut all three off together. The internet, because it carries commerce, health, and communication through the same wires, made a new kind of instrument possible: switch the network itself off, and every use of it, wanted or unwanted, stops at once.

Governments began reaching for that switch early in the internet's political life. Egypt's government ordered the country's internet providers to sever their connections in the closing days of January 2011, cutting most of the nation's access for several days during the uprising that ended President Hosni Mubarak's three decades in power, an episode widely cited as one of the first national-scale demonstrations of the tactic. What has changed since is not the existence of the tool but its frequency. Researchers who track the practice, among them the KeepItOn coalition run by the digital rights group Access Now, now log the year's shutdowns the way a weather service logs storms: by country, by duration, by stated cause.

The definition trackers use is deliberately broad. A shutdown can mean a full network blackout, a block on specific applications such as messaging or social platforms, or a deliberate slowdown of connection speeds to the point that ordinary use becomes impractical. What unites them is intent: an order, usually from a government or a regulator acting on its instruction, that degrades or removes internet access for a population, a region, or a platform, for a stated reason that is almost always public order, national security, or the integrity of an examination.

The legal basis for a shutdown is almost always domestic. Most governments that order one point to telecommunications law that lets a minister or a home department suspend services in the interest of public safety or during a declared emergency, provisions written decades before anyone anticipated that the internet, rather than a single radio transmitter, would be the thing being suspended. Internet service providers, private companies operating under a government license, then have little practical choice but to comply, which is part of why the shutdown is a tool available almost exclusively to states: cutting a network requires control over the license, not control over the content running across it.

The price of silence: from emergency measure to routine expense

In 2016, the Brookings Institution's Darrell West made one of the first attempts to put a number on the practice, studying 81 short-term shutdowns across 19 countries between July 2015 and June 2016. His finding: those blackouts, most lasting a day or less, cost the global economy an estimated $2.4 billion, a figure built from the GDP lost in the affected regions during the days their networks were down.

A decade later, the picture bears little resemblance to that baseline. Top10VPN, a research group that publishes an annual accounting of shutdown costs, estimated that government-imposed internet shutdowns cost the global economy $19.7 billion in 2025 alone. Set beside Brookings' $2.4 billion reading of the 2015 to 2016 period, that is a rise of close to eightfold in ten years, measured in a single year of shutdowns rather than a two-year survey.

The rise is not simply better counting. Access Now's KeepItOn coalition recorded 283 documented shutdown incidents across 39 countries in 2023, the highest count in the years it has monitored the practice since 2016, and a 41 percent increase over 2022. The incidents are also lasting longer: shutdowns running five days or more made up about 15 percent of all incidents in 2022 and rose to more than 41 percent of them in 2023. A tool once used for a tense afternoon is now being used for a tense month.

The economics behind the multiplier are straightforward. A modern economy runs payments, logistics, retail, and services through the same connection a government can order cut, so an hour offline is not lost browsing time, it is lost transactions, unfilled orders, and idle call centers. The bill scales with how deeply a targeted economy has moved online, which means the more digitally developed a country becomes, the more expensive its own government's kill switch gets to pull.

India: the routine instrument

No government has reached for the switch more often than India's. Access Now's KeepItOn coalition recorded 116 documented shutdowns in India in 2023 alone, extending the country's run atop the global shutdown count to six consecutive years. The same report estimated the toll of that year's shutdowns at $1.9 billion, with $118 million of it in lost foreign investment recorded in just the first half of 2023.

The Software Freedom Law Centre India, which has tracked government-ordered shutdowns since the practice became common in the country, counts 805 of them since 2012. More than half, 433, were imposed in the single region of Jammu and Kashmir, a pattern that marks the territory less as an occasional site of unrest than as a place where the internet is treated as a switch the state keeps a hand on.

The longest single instance illustrates the point. Following the government's August 2019 revocation of Jammu and Kashmir's special constitutional status under Article 370, authorities kept the region's internet cut for 552 days, from 4 August 2019 to 6 February 2021, at the time the longest sustained regional internet shutdown recorded anywhere in the world. A blackout that outlasted many national election cycles was applied not as an emergency response to a single event but as a standing condition of the region's governance.

What India's record shows is that a kill switch, once normalized, stops functioning as a last resort and starts functioning as a first one. Shutdowns have been ordered ahead of school examinations to prevent cheating, during festivals and protests to manage crowds, and around elections to slow the spread of rumor, uses that sit well outside the acute national-security justification the tool was first associated with. The frequency is the story as much as any single incident: a democracy with hundreds of documented shutdowns has folded the switch into its ordinary administrative toolkit, not kept it in reserve for emergencies.

India's courts have pushed back, with limits. In January 2020, the country's Supreme Court, ruling on the Kashmir shutdown in Anuradha Bhasin v. Union of India, held that an indefinite suspension of internet access was not permissible and that shutdown orders had to be temporary, reviewed periodically, and made public rather than issued in secret. The ruling did not end the practice; India's shutdown count continued rising in the years after it, including the 116 recorded in 2023. It did establish, at least on paper, that the switch could not be thrown and simply left off, a distinction that separates the courts' tolerance of the tool from the government's appetite for using it.

Iran, 2026: the crisis instrument and the longest blackout on record

If India shows what the kill switch looks like as routine governance, Iran's 2026 blackout shows what it looks like at full scale. The shutdown began on 8 January 2026 amid protests, then deepened after 28 February 2026, when American and Israeli strikes hit Iranian territory and the country's connectivity collapsed to roughly 1 percent of its normal traffic, according to NetBlocks monitoring cited in reporting by Al Jazeera. The blackout did not lift until 26 May 2026, 88 days after it began, the longest nationwide internet shutdown recorded by any country to date, per reporting by TechTimes.

The daily cost was correspondingly severe. Iran's own Chamber of Commerce estimated the blackout's direct economic losses at $30 to $40 million per day, a range that, sustained across 88 days, implies losses running into the billions before any account is taken of the businesses that closed, the remittances that could not clear, or the exporters cut off from buyers who could not reach them.

The shutdown's onset shows the two models blending. Its first weeks in January looked like a scaled-up version of the domestic-control shutdowns seen elsewhere: intermittent restrictions layered on top of a protest movement, similar in kind, if not in duration, to blackouts governments have used to blunt unrest for over a decade. The near-total collapse that followed the February strikes was a different order of event, a wartime posture with little real precedent at this length, and it is that second phase that produced the 88-day record and the bulk of the estimated economic loss.

The blackout's two phases are worth separating. The first, tied to January's protests, resembled the tool India uses most often: a government narrowing communication to blunt organized dissent. The second, tied to wartime strikes at the end of February, was closer to a wartime communications posture, an attempt to deny an adversary situational awareness and deny the country's own population unfiltered information about the strikes, at a cost the government's own commerce body put in the tens of millions of dollars a day.

Together the two phases make Iran's 2026 blackout a case study in how far the instrument can be pushed once a government decides the value of denying information outweighs its price. Eighty-eight days is long enough that the blackout stopped functioning as an interruption to Iran's economy and became, for that stretch, closer to its baseline condition.

Myanmar: a coup's daily bill

Iran and India show the tool used for long duration and for high frequency. Myanmar's 2021 coup shows what a single, decisive break can cost within a year. When the country's military seized power on 1 February 2021, it moved quickly to cut internet access and block social media platforms nationwide, tactics it repeated through the months of protest that followed.

Analysis compiled by the Internet Society's Pulse initiative put the economic toll of that year's shutdowns and blocks at an estimated $2.8 billion for 2021 alone, described as the single largest shutdown-driven economic loss recorded anywhere in the world that year. NetBlocks, whose monitoring underpins much of the shutdown-tracking field, separately estimated that a single day-long nationwide outage during the coup period cost Myanmar's economy roughly $22 to $24 million, a per-day figure in the same range Iran's Chamber of Commerce would later estimate for its own 2026 blackout.

The similarity is instructive. Two governments, five years apart, arrived independently at internet shutdowns as their response to a moment of acute political danger, a coup in one case and wartime strikes combined with protest in the other, and both paid a bill in the tens of millions of dollars for every day the switch stayed off. A tool that different regimes reach for under different pressures, and that costs roughly the same order of magnitude wherever it is used, has stopped being an improvisation. It is now a known, priced option on the shelf of state power.

What goes dark stops being read

Every government that has ordered a shutdown has offered a version of the same justification: public order, national security, the prevention of violence coordinated over social media, or the integrity of an examination. There is a genuine case on that side of the ledger. Communication networks have been used to organize riots, coordinate attacks, and spread disinformation at speed, and a state that believes it faces an immediate threat to life has a real interest in the tool, however bluntly it is applied.

The case against it is just as concrete, and it is now priced in dollars rather than argued in the abstract. Every shutdown blocks the organizing its architects fear alongside the ordinary commerce, healthcare coordination, education, and journalism the network also carries, indiscriminately, for as long as the switch stays off. Access Now's KeepItOn coalition, along with United Nations human rights bodies, has argued for years that shutdowns rarely achieve the narrow security aim claimed for them and instead impose broad harm on the very population a government says it is protecting. The rising bill, from $2.4 billion a decade ago to $19.7 billion in the most recent year measured, has not made governments use the tool less. It has made the tool more visible, and its true cost harder to deny.

There is a newer cost too, one this publication's own subject matter makes visible. A country or a region cut off from the internet does not simply lose access; for the duration of the blackout, it also stops being crawled, indexed, and read by the systems that now decide who gets found, cited, and named. A search engine or an AI answer engine can only surface a business, a hospital, or a government service it can reach, and a blackout is a period during which nothing behind it can be reached at all. Jammu and Kashmir's 552 days, and Iran's 88, were not only periods when residents could not get online; they were also periods when the outside world's machines could not see in.

The cost estimates themselves rest on a fairly consistent method across trackers: take a country or region's normal internet-dependent economic activity, approximated from measures such as internet penetration, mobile connectivity, and the share of GDP that depends on being online, and calculate what a day, a week, or a year without that activity removes from output. It is a modeling exercise, not a ledger of receipts, which is why serious trackers label their numbers as estimates, and why independent readings of the same event, Myanmar's NetBlocks figure and Iran's Chamber of Commerce figure among them, land in similar but not identical ranges. The method's limits do not undercut its central finding: whichever way the estimate is built, the number has moved in only one direction over the last decade, and that direction is up.

That is the throughline across a decade of rising shutdown costs. Every age's dominant medium has rewarded whoever controlled it and cost whoever was cut off from it, and the internet has simply made both the reward and the cost easier to measure, and easier for a government to invoke at will.