WHEN LU WOKE up early on February 1, 2021, he noticed that his Wi-Fi had stopped working. He turned his router off and on again. Nothing changed: there was no internet. Worse, it now looked like his smartphones couldn’t get online either: Facebook’s app loaded endlessly, WhatsApp texts stayed stuck in pre-delivery limbo. Lu stepped out onto the balcony of his flat and spied the glimmer of a TV screen through his neighbour’s window. Lu shouted, asking whether he knew what was going on. “It’s strange,” the neighbour shouted back. “On the news, they officially announced that the military took over the country.”

This is how Lu, a trade union worker based in Myanmar’s capital Yangon who asked to be identified only by his first name for security reasons, realised that his country’s democracy had been upended by an army coup d’état. For months, Myanmar’s armed forces had been questioning the legitimacy of the November 2020 general election, which had resulted in a landslide for the incumbent governing party, the National League for Democracy, and its leader, Nobel Peace Prize laureate Aung San Suu Kyi. On February 1, hours before the new parliament was due to be sworn in, troops swooped on Yangon, arrested Aung San Suu Kyi and other political leaders, and installed bespectacled general Min Aung Hlaing as the country’s ruler.

In an act of textbook military subversion, the coup played out in darkness: at night, and amid a systematic information blackout. Monash University’s IP Observatory, an academic project based in Australia that monitors global internet traffic, first clocked a decrease in Myanmar’s connectivity around 1am local time. Another such project, Internet Outage Detection Analysis (IODA) at the University of California San Diego, detected a substantial slump around 3am. The shutdown was piecemeal and imperfect: different internet service providers discontinued network services at different times, and the loss of signal was not uniform across the country – the southernmost Tanintharyi region, for instance, remained unaffected. This pattern makes sense if one pictures jeep loads of soldiers barging into the offices of Myanmar’s telecom companies at different hours, demanding the shutdown at gunpoint.

Judging by a further collapse in connectivity registered by IODA, by 6:50am most of Myanmar was offline. Voice and text cellular services were switched off across the country and the national broadcaster went off-air. As armoured vehicles rolled down the streets and military checkpoints popped up along thoroughfares, Myanmar’s population was left blindsided.

“We couldn’t communicate with each other anymore, and people totally freaked out. The entire country was blocked,” Lu says. He and his wife idled time away at home, waiting for something to happen. “Nobody dared go out, nobody knew what was happening on the next street, or what was happening in another township.”

Connectivity began to slowly and patchily inch up around 10am. Lu says that his Wi-Fi connection came back earlier than mobiles data, and that he got online earlier than friends living in other parts of the country. But things were not back to normal. Over the following days, thousands of citizens took to the streets and social media to orchestrate protests and denounce the military’s power grab. The junta’s response was to come down hard on protesters, in the form of physical violence and digital muzzles.

Mass arrests and nightly raids against dissidents multiplied; journalists and democracy watchdogs were intimidated; security forces turned water cannon, tear gas and rubber bullets against peaceful protesters waving three fingers as a gesture of defiance. Live ammunition would follow. On February 4, Telenor Myanmar – a subsidiary of a Norwegian telco, and the country's most popular mobiles operator – announced that the junta had ordered mobiles operators and internet service providers to block access to Facebook, which protesters had been using to organise. A day later, Instagram and Twitter were blocked. Following that, the internet went dark again from the early morning of February 6 until the afternoon of February 7. This time, orders from the telecom regulator reached internet service providers in unison, and the drop in connectivity was synchronised and all-encompassing – traffic coming out of Myanmar was snuffed to negligible levels.

This turned out to be a dress rehearsal: on February 15, Myanmar experienced the first of a long string of internet curfews. Every night at 1am until 9am, the country was cut off from the global internet – “as if swallowed by a snake”, Lu says. Simon Angus, who works as a researcher at the IP Observatory, describes it as “a precise, metronomic shutdown.”

On March 15, the mobiles internet was switched off completely; on April 2, wireless broadband was discontinued. Those who still managed to connect via fixed broadband were confronted with a husk of the network they remembered: sluggish, heavily censored and hard to navigate without relying on software such as virtual private networks (VPNs).

For people like Lu and his wife, who works as a software developer, it was a professional catastrophe. “She used to have late meetings, even at midnight – now she cannot work, she can’t go to meetings,” he says. Lu knows university students who used to attend lessons remotely amid the Covid-19 crisis, and now feel educationally stranded; he knows food delivery workers who, with no mobiles internet, cannot take orders. “A lot of people lost their jobs because they needed to work with the internet,” he says. Even though the curfew ended and fixed-cable connectivity was reinstated on April 27, mobiles and fixed wireless remained inaccessible as of June 2021.

There is a grislier element to the shutdown, in the wider context of brutality and repression. A Burmese human rights activist, who asked not to be named for fear of retaliation, is filled with dread whenever her connection dies. “This internet shutdown doesn't happen in a situation where everything is going well,” she says. “They are cracking down on the protests, killing civilians. You live in fear that something can happen to you at night. And you think: if there is no internet, you cannot talk to your friends or colleagues about what is happening.”

THE SITUATION IN Myanmar is not unique. Other countries, including India, Ethiopia, Belarus and Venezuela have used internet shutdowns as a method of crushing dissent and obfuscating the truth. The internet as a tool of knowledge-sharing, connection and bottom-up democracy is being bludgeoned with an off-switch.

The first major precursor to today’s internet shutdowns was the Arab Spring. Starting in December 2010, as people across North Africa and the Middle East rose against authoritarian rulers, the internet emerged on the world stage as a force for political mobilisation. Protest movements were hatched in Facebook groups, spotlighted on Twitter and chronicled on YouTube, regime brutality and extrajudicial killings included. English-speaking newspapers started talking of “Twitter revolutions”.

How key a role social media played in the turmoil – which touched over ten countries, brought down four dictators, triggered at least two civil wars and destabilised the area to this day – is a matter of debate, but it seems clear that some beleaguered tyrants felt threatened by it. On January 27, 2011, two days after protesters had started amassing in Cairo’s Tahrir Square, Egypt’s Hosni Mubarak took the country offline, and continued to block services for five days. In Libya, Colonel Gaddafi followed suit, reacting to a budding uprising with a barrage of internet disruptions culminating in a four-day-long shutdown on March 3. The move drew international condemnation, prompting the United Nations to speak up against the suppression of freedom of speech.

For the dictators, it arguably backfired. Jared Cohen, a former US State Department staffer who had gone on to join Google’s “think-do tank” Google Ideas and was in Egypt during the revolution, said in a July 2011 interview that Mubarak’s shutdown had motivated many young Egyptians, who may never have joined the protests otherwise. By the end of the year, former president Mubarak would be standing trial in a Cairo courtroom and Gaddafi would be dead.

Access Now, a digital rights advocacy group founded in 2009 in the aftermath of a spate of internet disruptions in Iran, has been keeping a tally of internet shutdowns from 2016 onwards. Put the numbers on a chart, and you can see a line hopping upwards year after year. In 2016, Access Now reported 75 shutdowns; in 2017, 106; in 2018, 198; in 2019, 213. In 2020, for the first time in half a decade, the group noticed a decrease from the previous year, down to 155 internet shutdowns in 29 countries. But, once you remember that 2020 was a pandemic year – a period of lockdowns, working-from-home, and extreme reliance on the internet to do everything from buying groceries to attending school – that number nevertheless seems enormous. “It was surprising to see governments – and we continue to see them – shutting down the internet during a time when, increasingly, we have all realised the importance of staying connected,” Felicia Anthonio, a campaign co-ordinator at Access Now based in Ghana, says.

The rise in confirmed shutdowns may partly be a consequence of increased awareness and reporting. Observers and diplomats, however, believe that turning off the internet has become an increasingly common tactic. “It's really a crisis for freedom of expression in many respects. And it's definitely a crisis that's expanding around the world,” says David Kaye, a professor of law at University of California, Irvine and the UN special rapporteur on freedom of expression between 2014 and 2020. “What’s worse, this tool is becoming almost normalised, even in places where the rule of law really should exist.”

Although shutdowns might seem a hallmark of autocracy, the worst offender is the world’s largest democracy: according to Access Now, India accounted for 109 internet shutdowns throughout 2020 – all of them enforced at a local level, such as in the disputed Jammu and Kashmir region, which went through 18 months of on-off shutdown. The record-holder for the longest shutdown ever is Myanmar, which kept several towns in Rakhine and Chin State – two northern states torn apart by ethnic violence and armed clashes – offline for 19 months; in February, the junta made a point of putting an end to the situation, and then unplugged the whole country.

Most governments do not acknowledge that a shutdown is underway, variously blaming the disruption on technical problems or foreign cyberattacks, or simply lying about the reality on the ground. When a rationale is provided, shutdowns are described as last-ditch measures to prevent violence, defend national security and, increasingly, to stem the spread of “fake news” online. (On February 6, Telenor cited the “circulation of fake news, stability of the nation and interest of the public” as the Myanmar government’s legal ground for the shutdown.)

There is a genuine debate to be had about how much power governments in poorer countries have when asking Silicon Valley giants to more proactively counter propaganda, hate speech and disinformation targeting their citizens. A western diplomat who was in Ethiopia in June 2020 – when the government imposed a three-week shutdown after the murder of a musician sparked riots and ethnic conflict – says that the risk coming from Ethiopian diaspora members calling for violence on social media was real.

Yet, Anthonio says, there is no evidence that shutdowns have a positive effect against disinformation or internet-incited violence. Quite the opposite. Myanmar’s military is infamous for running disinformation campaigns on Facebook encouraging genocidal violence against the Muslim Rohingya minority; following the February coup, it kept flooding Facebook – theoretically blocked in the country – with falsehoods, until the social network booted military-linked accounts from the platform. “What they wanted to accomplish was getting people not to see the kind of brutality that was going on around the country,” says Harish Nair, managing editor of FactCrescendo, an Indian fact-checking outlet that also covers Myanmar. “They wanted to add their own ‘fake news’, and show that pro-democracy supporters are causing chaos and commotion everywhere.”

It is telling that, alongside mass protests, the kinds of events that seem to be most associated with internet shutdowns are elections. A blocked or limited internet in the lead-up to an election means that the opposition will be less able to co-ordinate, canvas and campaign; no internet during an election means that irregularities won’t be immediately reported; and no internet in the aftermath of a rigged vote makes it harder for citizens to voice their discontent. In certain African countries, election-day shutdowns have become so commonplace that internet-watchers put reminders in their calendars.

Doug Madory, director of internet analysis at network analytics company Kentik, has been monitoring shutdowns for over a decade. He had expected that, after the disasters of Egypt and Libya, full-out shutdowns would be replaced by less disruptive strategies – censorship, blocking, targeted removal of websites. But many of the countries implementing shutdowns do not have the experience or capabilities to act more elegantly. Myanmar had just over 495,000 internet users in 2011; now it has over 20 million. Across Africa, the percentage of people connected to the internet nearly trebled from 2010 to 2019. Further complicating the situation is the fact that encrypted messengers, VPNs and other tools for bypassing internet restrictions have become more widely available, prompting governments to come down heavily. “It seems like we've regressed a little,” Madory says. “Or maybe we never progressed.”

"THE INTERNET WASN'T designed with measurements in mind,” says Alberto Dainotti, a research scientist and founder of the internet outage tracking project IODA. The internet is a kludge – an effective but tangled network grown through spontaneous expansion and shaped by makeshift fixes ossified into structural features. There’s no simple system to track whether a government has taken a country offline. Observers such as IODA have to come up with proxies, ways to detect the fire of an internet shutdown by spotting a sort of digital smoke.

One of the most rudimentary methods to enforce a shutdown is called BGP manipulation. BGP, or Border Gateway Protocol, has been variously described as the internet’s post office, satnav or air traffic control tower. It is the system that allows a packet of information to find its way from point A to point B through the tangle of nodes and connections making up the internet. Through BGP, every node in the network – representing the entry point for a group of internet addresses, which in turn represent users – constantly advertises which addresses it gives access to. These announcements are picked up by the most immediate neighbours and echoed through the whole internet, blazing a trail between two points in the network. This is how a user in one locations can load a website from a server based in another locations. Governments can order whoever is in charge of the nodes advertising entry to a cluster of addresses – typically an internet service provider or telecom company – to withdraw those announcements, effectively wiping those users off the global internet map.

Withdrawing BGP routes is convenient as, to an extent, it can allow for select groups of addresses – those belonging to government officials, say – to retain their internet connection while everyone else is in the dark. Cutting the internet cables or powering down the country’s networking hardware – the crudest methods to carry out a shutdown – could also risk affecting other countries’ connectivity. One Syrian telecom specialist, who was employed by government-owned Syrian Telecom throughout several shutdowns that took place in the country from 2011 onwards, says he was asked to use “BGP techniques” to ensure that a few select people could still get online while citizens protesting against the al-Assad regime were cut off.

The specialist, speaking anonymously out of security concerns, recalls removing every set of addresses from the company’s BGP announcement, except for those linked to what he calls “VIP customers”. In one sense, he considered the technique a way to put a target on those people’s backs. “Anyone who looked to Syrian Telecom’s BGP table [would have seen that] it just contained VIP customers’ [IP addresses],” he says. “It was like a message to hackers and researchers: ‘please attack them!’” He used BGP manipulation again when Syria started enforcing nation-wide internet shutdowns to prevent cheating during school finals. (This practice has become popular elsewhere, including in Iraq and Ethiopia.)

For a government, one downside of BGP manipulation is that it can be easily detected by researchers. Its targeting is also rough: addresses are taken offline in gross bunches. Firewalls – systems able to filter inbound and outbound traffic – are a more refined alternative, able to narrow their scope down to the individual address if necessary. They are usually built by grafting devices known as middleboxes onto a network’s cables. Ramakrishna Padmanabhan, a researcher with IODA, explains that middleboxes can be programmed to block all traffic going to a specific destination address or coming from a particular source address, or else entire internet protocols or classes of content such as video, voice or email. “It's a fairly computing-intensive process,” Padmanabhan says. Middleboxes also have legitimate applications, which makes it harder to control exports of this technology.

Middleboxes can block access to individual popular websites – a move that does not qualify technically as a shutdown, but the impact of which can still be staggering in countries like Myanmar, where much online activity takes place on Facebook. Done with systematic completism – with the exclusion of all social networks, search engines, news websites, app stores and video sharing platforms – this method can be as effective at disconnecting a country’s populace as sawing off a submarine cable. The blocking of specific sites can theoretically be circumvented by using VPN apps, which tunnel traffic through servers in other countries, but governments can also block websites where these apps are downloaded, or try to identify and winnow out VPN traffic.

Firewalls are subtler than BGP manipulation: to verify filtering by firewalls, IODA uses “active probing” – essentially pinging networks known to be at a certain geographic locations. Most networks are designed to automatically respond to pings by echoing them back to the sender, unless something prevents them from doing so. No pong after the ping indicates that a system may be firewalled.

Except, Dainotti says, most networks are not all of them: it is perfectly possible for a network to be connected to the internet and yet not return a ping. To verify, IODA uses pings in concert with BGP observations and another proxy. Called a “network telescope”, it is a router at the University of California San Diego that continuously advertises itself as open for internet traffic. Inevitably, it receives a deluge of unsolicited garbage. “A large part of it is malware,” Dainotti says. “A large part of it is traffic sent to it due to misconfiguration. A large part of it – we found out in 2016 – was BitTorrent clients mistakenly sending traffic to our offices.” That spontaneous stream of uninterrupted pollution – spread by infected computers, botnets scanning the web for prey and accidental connections – is known as “internet background radiation”, and it can be parsed to different geographic locationss. A drop in the amount of radiation emanating from a specific country suggests that a shutdown may be occurring.

Other internet measurement outfits contribute other techniques. International nonprofit OONI (the Open Observatory of Network Interference) detects the blocking of specific websites and messaging services: users download OONI’s app to confirm whether it is impossible to connect to a given website from where they are. From his listening post at Kentik, Doug Madory pores over aggregate data to spot shutdowns, including at the level of the mobiles internet – which is harder to pick up with pings and BGP tables. The IP-Observatory in Australia pings millions of internet-connected devices across the planet to pinpoint shutdowns and gauge internet throttling – a variation on the shutdown theme in which connection speed is toggled down dramatically, thus transforming internet navigation into an infernal slog.

In 2016, Access Now spearheaded the creation of the Keep it On coalition, a loose alliance of over 240 groups collecting and sharing information about shutdowns, and providing assistance to the people affected by them. Their goal is keeping up the pressure on governments, through courtroom battles, awareness campaigns and technical aid to people on the ground, so that internet shutdowns are taken off the table as a political tool.

“In the long term, in my ideal world, I would love to see governments themselves recognise the need to stop using shutdown against their own citizens and rather invest in internet infrastructure and ensure that the internet is accessible for everyone,” says Access Now’s Anthonio. “But that is the ideal scenario.”

SIMON ANGUS, THE IP Observatory researcher, compares internet measurement to the detection of illegal nuclear tests. “You have various agencies measure different telltale signs of a nuclear explosion – some will focus on sonic booms, others on earthquakes, others on radiation and so on – and then they triangulate.”

While the parallel may seem over the top, many researchers think about internet shutdowns as an arms race. Shutdowns pit governments against their citizens, with the latter constantly taking up tools to try to circumvent disruptions. When governments feel they’re losing the battle, they hastily turn to the nuclear option – just shutting everything down.

But, in the long run, this is not a tenable position. Shutdowns are painful and very expensive: even if you don’t end up like Mubarak or Gaddafi, you might still end up with an economy in tatters. A Brookings Institution study estimates the global cost of the 81 shutdowns carried out worldwide between July 2015 and June 2016 at $2.4 billion. In the age of real-time communication, virtually no business can hope to thrive without reliable internet. In January 2020, a Reuters story described how thousands of Kashmiris – left jobless after the shutdown devastated the region’s tourism industry – travelled every morning on a train known as the “Internet Express” to towns unaffected by internet restrictions, in order to fill out online job applications and check business emails.

Belarus, once a thriving startup scene, is now haemorrhaging tech founders after president Alexander Lukashenko wielded shutdowns to cling to power amid mass protests in summer 2020. One Belarusian technology entrepreneur, speaking anonymously out of security concerns, says that, in the wake of the shutdown – and of the government-sanctioned violence – Belarus’s reputation as a high-tech hub was “just destroyed”. “The whole brand of Belarus as a great place to build software or set up a startup’s office – that evaporated, it was gone,” he says. “Most of the highly-skilled professionals, be that engineers, or product managers, or marketers, packed their suitcases and left.”

This is why the Myanmar military kept what little was left of the internet running during business hours, and why business districts are usually spared from shutdowns in India.

Over time, governments may consider alternatives to the prolonged curfews or outright outages seen in Myanmar, Belarus or Ethiopia. Some researchers believe that shutdowns may actually become less frequent, as more countries move towards creating a permanent internet-censorship apparatus instead of blocking access every now and then. Their ideal end-state would probably look somewhat like Iran or China, two pioneers of internet censorship.

Iran has perfected the art of severing itself from the global internet while maintaining internal connectivity by creating the National Information Network, a country-wide intranet that lets everyday activities go on relatively unperturbed on Iran-built, surveillance-ridden apps while silencing critical voices. The network was put to the test during a weeklong shutdown in November 2019, as the country was awash with protesters. “Iran is smart: they wanted to kill the protesters, but they didn't want to anger people even more,” says Iranian digital rights activist Amir Rashidi. “Which is why when they shut down the internet, the local net kept working.” (Russia has developed a similar domestic network but has only unplugged from the global internet during a short pilot trial in 2019.)

And China, of course, has spent almost two decades perfecting its Great Firewall, a prodigy of fine-tuned censorship technology that rejects the internet as a venue for self-expression and political action while harnessing its potential for business and totalitarian control.

Shutdowns, then, may be only the first step on the path. “You start by going into the tool shed and you pull out the large axe,” Angus says. “Then eventually, you realise, ‘Maybe I don't need an axe, maybe I need something more like a screwdriver, or a watchmaker’s tool.’ You can have suppression – but not crush the economy.”

Already China is exporting its model to other countries. In November 2020, the US Treasury imposed sanctions on Chinese state-owned electronics manufacturer CEIEC for selling a commercialised version of China’s Great Firewall to Venezuela, a serial shutdown perpetrator. Others may follow: reports in early May 2021 suggest that Myanmar’s military is planning to create an “intranet”. Shutdowns mostly affect countries that only recently experienced massive internet adoption, and which are now at a crossroads between competing conceptions of the internet. The risk now is that millions of budding internet users may end up using something that is very little like the internet most countries know today.

One future path, then, could see the age of the shutdown progress into the age of the splinternet – the fragmentation of the global internet into multiple national networks held in the vice grip of their respective governments. “We're headed for the balkanisation of the internet,” says Harry Halpin, a visiting professor at Belgian research university KU Leuven. “And this will hurt the internet as a space of universal communication and information-sharing, which has saved many people's lives. That still is, I think, a thing worth fighting for.”

• 🌎 Sign-up to WIRED’s climate briefing: Get Chasing Zero
• Hidden Arctic caves hold secrets about our past and future
• Fed up of your job? You could join the Great Resignation
• Zoom dysmorphia is following us into the real world
• On YouTube, you’re never far from a dying kitten
• Too much Google? Try these Google search alternatives
• A radical plan to treat Covid’s mental health fallout
• The 100 hottest startups in Europe in 2021
• 🔊 Subscribe to the WIRED Podcast. New episodes every Friday

This article was originally published by WIRED UK