Cars, Speakers, TVs, Phones… Just How Many Devices Around You Can Record Your Every Move?

No urban legends here — every eavesdropping case below is backed by evidence.

How do you get complete privacy? If you follow Edward Snowden's method, you check into a hotel and then stuff a pillow against the gap under the door, unplug the room's landline, tape over your laptop's webcam, and shut your phone inside the minibar fridge.

That is what Glenn Greenwald, the Guardian columnist who helped expose the PRISM program, watched Snowden do when the two first met in Hong Kong. Like many people who met Snowden afterward, Greenwald took the behavior at first for excessive paranoia.

Privacy attracts urban legends: that every parcel gets opened, that a phone left switched on can be turned into a bug from a distance, that the cameras on the street know who you are.

The countermeasures seem plentiful too. Encrypted messaging tools, VPNs and Tor are none of them hard to get.

But set the unproven legends aside and look only at cases backed by hard court documents, and you find that today's surveillance reaches far past what is stored on your phone or your computer. Government agencies need none of the 007 tricks Snowden was guarding against in that hotel room. Commercial companies help them do it easily.

The latest device to be exposed is the car. Information released by US judicial authorities shows that several in-vehicle services, including the OnStar system used by General Motors, have been helping the government track owners' locations and listen in on conversations inside their cars for more than a decade.

Your car can give away your location and record what you say in it — and even the radio can log where the car has been

Late in 2006, the US Drug Enforcement Administration learned from an informant that a dealer named Lamauro Coleman regularly drove from Michigan to New York to pick up heroin. The informant had twice bought more than 100 grams of heroin from Coleman himself.

In January the following year, when Coleman drove to New York for another pickup, the informant tipped off the DEA. Several agents tried to tail his car, but Coleman was good at spotting them and shook them off again and again.

Coleman happened to be driving a GMC Envoy SUV fitted with the optional OnStar system, so the DEA tracked the vehicle straight from the system's back end. He was pulled over and arrested in Michigan.

OnStar, owned by General Motors, is an in-vehicle communication service of the same kind as Toyota's G-Book. It offers route advice, emergency rescue calls, remote diagnostics, vehicle location and more.

Technically, OnStar's built-in positioning system knows where the car is, so that when a driver presses the alarm button in an emergency, rescuers can reach the scene faster.

Using it is like using a human version of Siri: press the call button and you are connected to a service center far away, where an operator plans your route or calls for rescue based on what you say.

So a car with a service like this has, in principle, a microphone in the cabin, a data connection and positioning. In Coleman's case, law enforcement simply asked OnStar to cooperate with police and took his location from the back end.

That same year brought a stranger OnStar case. In 2007, the owner of a Chevrolet Tahoe in Ohio was arrested by police for hiding marijuana. He had not paid for the OnStar service, so although the hardware was built into his car, the service was not supposed to work. But after he pressed the OnStar emergency call button by accident, an OnStar operator who should never have been connected overheard him discussing marijuana with someone else and called the police. Officers stopped his car on the strength of what OnStar told them.

Court documents record that as early as 2001, another connected-car provider, ATX, was ordered to hand over audio from a Mercedes-Benz S430V. That November, ATX ran 30 days of voice surveillance on the target vehicle at the FBI's request.

In other words, the government has been watching citizens through their cars for 15 years.

IHS data shows that more than 11 million cars with connected systems were sold in the United States in 2015. By 2021 the figure will reach 16.5 million — meaning nearly every new car will have the feature. Beyond OnStar and G-Book, smart in-car functions keep multiplying. Audi, Mercedes-Benz and BMW made crash location and emergency calling standard long ago, and the principle is almost identical to OnStar's. Ford's Sync 3 and Honda's Honda Connecting have no emergency call service, but because they need traffic data they come with a 4G connection built in, plus voice recognition — so in principle they too could listen and track.

And even if the owner leaves home without a phone and the car has no connected system like OnStar, the radio alone can log where you are.

In 2014, at the request of New York police, the satellite radio provider SiriusXM used the satellite radio unit inside a car to track a Toyota 4Runner SUV suspected of links to an illegal casino. Satellite radio has no positioning of its own, so police fixed the car's location by triangulation, much the way phones are located. SiriusXM admitted it had granted five similar police requests, but it would not say whose.

All of this happened in the United States. Not every country discloses law enforcement's surveillance methods in court the way the US does.

After PRISM, even your speaker listens to you all day

Edward Snowden exposed PRISM, a US government program that carried out mass surveillance of user communications through internet companies. Most of what his documents showed, though, was how much data the government collected, how the big companies cooperated, and roughly whom it targeted. Only when the first case of PRISM being used against a named individual came to light did the program's real menace become concrete.

Sometime before 2012 or 2013, Tony Pullman, a New Zealand citizen born in Fiji, was suspected by New Zealand law enforcement of plotting to overthrow the government. His offense was posting comments on social media against Frank Bainimarama (Josaia Voreqe Bainimarama), the military leader who had seized power in Fiji in a 2006 coup and was then running the country.

New Zealand agents worked with the US National Security Agency and used PRISM to collect a great deal of Pullman's personal information: his Gmail messages, his bank statements, geotagged Facebook photos, his chat logs. His home was then searched, his passport was cancelled, his company was placed under surveillance. In the end it turned out that Pullman had done nothing at all. He had said a few things on the internet.

Anyone with a little counter-surveillance sense knows that anonymizing tools — a VPN, a browser's private mode — will shake off routine tracking online. But surveillance can also come from directions you would not expect, and not through your computer.

In the summer of 2015, Amazon began selling the Echo, a smart speaker with a voice assistant that answers in real time. Say "Alexa" at any moment and you can give it an order: play this album, read this ebook.

To do that, the Echo must analyze the sounds around it every second, so that it knows when you are calling it. In theory that work can all be done on the device, with nothing recorded or uploaded. But a murder case suggested the Echo may have been recording daily life.

On November 21, 2015, Victor Collins, a man from Arkansas, went with a few friends to James Bates' house at Bates' invitation to drink and watch a game. By the next morning, Collins was found dead in Bates' bathtub.

The other two friends left at midnight. Bates said he went to bed at 1 a.m. while Collins stayed up watching the game. With no witnesses and the victim dead inside a private home, Bates became the prime suspect. He insisted he was innocent and had no idea how his friend had died in his house.

Bates loved gadgets. His house had a Honeywell alarm system, a Nest thermostat and plenty of other popular smart-home equipment, along with an Amazon Echo. Late last year, as prosecutors prepared to try Bates for first-degree murder, they argued that the Echo might have picked up sounds from that night and demanded that Amazon hand over the user's recordings. Amazon's answer was not that no such data existed. It was a refusal to hand the data over.

Amazon said no — but the refusal suggests, indirectly, that the Echo is not only listening to you all the time. It may be recording you.

Every voice-controlled smart appliance carries the same risk. In 2015, Samsung's television terms of use stated plainly that sets with voice recognition would record users' voices and might send them to a third party. Samsung said the third party was in fact a search engine, but the arrangement made it easy to collect user data quietly.

Since the Echo took off, tech companies have treated voice control and voice recognition as an innovation worth chasing, and have shipped many products that work the same way.

Last October, Google launched the Home speaker, very close to the Echo in function and ready to answer a user's call around the clock. In China, JD.com and Singapore's Creative Technology have both released smart speakers that do much the same thing. iFlytek, a voice-recognition specialist, has one too.

Tom Clancy, the military and spy novelist, described a counter-surveillance trick used by the Soviet brass in Red Tide (《红潮》): officers taking a sauna together would throw cold water on the floor, and the hiss would keep listening devices from catching what they said.

Films are full of similar tricks — run the tap, turn up the television. Against today's microphones and computer speech recognition, they no longer work very well.

Take the Echo's automatic speech recognition system, called ARS. With an array of seven microphones and a processing system, it can pick a human voice out of a noisy room. In practice you can put on very loud music, call to it from five meters away, and it will still make out what you said.

A phone that can listen to the room at any moment is not an urban legend

At 33 Thomas Street in Lower Manhattan stands a 29-story building. It is the color of concrete, with no windows and no sign; at the ground-floor entrance there is an intercom. Every night, as the lights come on in the buildings around it, it disappears into the dark, and its huge vents give off a dull, rhythmic hum.

Completed in 1974 and designed by John Carl Warnecke & Associates, this is the Long Lines Building, one of New York's oldest and best-known towers, fortress-like from the outside. Ownership records say the 160-meter building — able to withstand a bomb attack, able to hold 1,500 people working sealed inside, with 29 floors above ground and three below — belongs to AT&T, one of the largest telecom companies in the United States, and is one of the country's largest long-distance call processing centers.

In Edward Snowden's documents, though, the Long Lines Building has another name: TITANPOINTE. According to Snowden it is not only an AT&T telephone center but one of the listening posts built in partnership with the National Security Agency to monitor phone calls, faxes and internet traffic. An investigation by The Intercept found that inside this building AT&T worked with the US government to spy on at least 38 countries and international organizations.

And more public court documents show that what a phone can hear goes beyond your calls.

As early as 2006, the FBI in New York, working an organized-crime case tied to the mafia, used a software flaw to push backdoor software onto a target's phone from a distance, then switched on the phone's microphone to listen to him. A phone already has a microphone and a data connection; once spyware gives someone control of it, it becomes an extremely good remote surveillance device.

A US district court ruled that remote eavesdropping legal, and it remains the first confirmed case of a phone being activated from a distance to listen in.

Back then you could still switch a phone off completely or pull out the battery, and with no battery no method of eavesdropping can work. Today almost every phone has a sealed-in battery, and sometimes switching it off does not help either. Wired has confirmed that certain phone trojans can wake a phone's communications processor periodically while the phone appears to be off, in order to listen.

In August 2016, Apple's iOS 9.3.5 update fixed a fatal flaw. The Israeli company NSO had been exploiting it, building it into eavesdropping software that could take complete control of an iPhone without leaving a trace — reading what was stored on the phone and recording the sounds around it.

Before it turned to the iPhone, NSO had spent years selling attack tools for Android; the government of Panama bought them for illegal wiretapping.

This is a rare case of a commercial company using a software flaw for targeted eavesdropping — but "rare" only means rarely disclosed. In an age of fast software iteration, flaws are certain to appear. The question is whether the software company can fix them before an attacker finds them.

Meanwhile, for the sake of a better voice experience, Apple, Google and Microsoft have all built voice assistants into their phones that wake at a spoken phrase. Leave the feature on, say "Hey Siri," and the iPhone starts waiting for your command.

Which means your phone, your tablet and your computer are listening to you all the time. According to what the tech companies have published, these assistants do their analysis on the device, the data is not uploaded, and there is no privacy risk. But the sheer breadth of the technology's deployment means that recording and recognizing speech 24 hours a day is technically feasible.

And several hundred million of the Android phones sold each year are beyond Google's control. What privacy protection they offer is anyone's guess.

A camera can do more than record video — it can also understand what you are saying

In many crime films the characters cover their laptop cameras to be safe. In an episode of Black Mirror broadcast last year, a hacker who had watched a man through his webcam while he browsed child pornography sites uses the footage to force him to rob a bank and kill, and the police catch him in the end.

Documents released by Edward Snowden suggest that being careful with your camera is not paranoia either.

They show that GCHQ, Britain's main signals intelligence agency, which answers directly to the Foreign Secretary, ran a surveillance operation called Optic Nerve in 2008, stealing data straight off backbone networks and violating the privacy of at least 1.8 million Yahoo users.

In the middle of 2008 the operation secretly captured and stored screenshots from video chats on more than 1.8 million Yahoo accounts worldwide. To keep the volume manageable, GCHQ saved one image per account every five minutes. The program also used facial recognition to pick people on watch lists out of the faces in those conversations, for closer surveillance.

The documents show that a great many explicit, intimate images turned up in the stored photos. A user's private video chat is supposed to be a safe and private space; this program left nothing of that privacy. Yahoo later issued a statement saying it had known nothing about the operation, that it was outraged, and that it would strengthen its data encryption.

Anyone who knows game consoles knows how much the Xbox Kinect motion-sensing camera can see: it recognizes faces, works out which controller belongs to which player in a multiplayer game, and uses infrared to analyze how a person's skeleton and muscles move. Snowden's documents show GCHQ was exploring surveillance through the Kinect camera on the Xbox.

There is no evidence yet that Kinect has actually been used this way, but if it were, it would probably see more than a laptop camera can: who comes into your living room, what they pick up, what they say. Even with no face in view it can identify someone by the habitual way the skeleton moves, matching motion to identity.

In 2001: A Space Odyssey, the protagonist suspects that HAL 9000, the all-knowing supercomputer running the ship, means them harm, so the men hide in a soundproofed pod to talk. HAL 9000 uses its camera to read their lips, learns what they are saying, and kills the second lead.

In Eagle Eye another supercomputer has gone rogue, and this time it can neither see faces nor hear sound. It works out what the protagonists are saying inside a safe room by reading the ripples in a glass of water on the table.

No one has reported the kind of sound analysis seen in Eagle Eye yet. The scene from 2001: A Space Odyssey has already come true.

DeepMind, the company behind the Go-playing AI AlphaGo, worked with Harvard University on a lip-reading system called LipNet, which reached 93.4% accuracy — better than any AI before it, and better than people.

For now the system still needs video of lips facing the camera. But at this rate it will be only a few years before the cameras that line the streets can not only watch everything you do but also know what you are saying.

Surveillance like this is nothing new

In the 1960s the Cold War between the United States and the Soviet Union spread into conflicts between and inside third-world countries, and the Vietnam War broke out. In 1967 the well-known news magazine Ramparts published a report revealing that the CIA was running a covert funding program at home, recruiting students from other countries at more than 400 American schools to do Cold War work.

After the story appeared, the CIA put the magazine, its reporters and its backers under illegal surveillance for years. A book published in 2008 confirmed that William Raborn, the CIA director at the time, personally asked for investigative reports on the magazine and its journalists.

The investigation into Ramparts was the first time in history that the CIA broke the National Security Act of 1947 by investigating a publishing organization. Things escalated quickly. Under a covert program later disclosed as "HT LINGUAL," postal workers piled up the letters and packages moving to and from the Soviet Union, China and other designated socialist countries in a room at Kennedy Airport every day. At night, agents opened the mail, photographed it, filed it, and then sealed it back up to look untouched.

Between 1953 and 1973 the program screened more than 28 million pieces of mail and opened 215,000 of them.

It was worse in the East. When Eastern Europe's political order collapsed in the late 1980s, the files of East Germany's internal security service — the Ministry for State Security, known by the contemptuous nickname Stasi — were opened in full, laying bare 40 years of indiscriminate surveillance.

The Stasi's slogan was "We are everywhere" (WIR SIND ÜBERALL), and by the results, they lived up to it. Its job was to support and carry out the Party's decisions; it acted as secret police and intelligence agency, and could exercise the full range of law-enforcement powers, prosecution, trial and imprisonment included. Between 1950 and German reunification in 1990 the Stasi recruited 274,000 full-time staff. They planted bugs in rooms, read mail, tapped telephones and otherwise gathered evidence of improper speech, then used threats and inducements to strip targets of their jobs and livelihoods, or to turn them into informants. The files record that the Stasi ran 175,000 informants and built secret dossiers on more than 6 million East Germans — more than a third of the population.

After reunification, the Stasi's headquarters and files were opened as a museum. Citizens who came to look up their own secret dossiers found that the people who had informed on them were their friends, their children, even their husbands and wives. The Lives of Others, which won Best Foreign Language Film at the 79th Academy Awards in 2006, showed the Stasi's methods and standard procedure in detail: bugs planted all over the room of a writer under watch, with staff assigned to listen 24 hours a day; pressure on his publisher until he lost his work and his income; his phone calls, travel records and mail monitored; his girlfriend threatened until she gave up what he said and where he went.

Today it is more efficient — and governments are no longer the only threat to privacy

The voice came from an oblong metal plaque like a dulled mirror which formed part of the surface of the right-hand wall. ... You had to live — did live, from habit that became instinct — in the assumption that every sound you made was overheard, and, except in darkness, every movement scrutinized. Winston kept his back turned to the telescreen. It was safer; though, as he well knew, even a back can be revealing.

1984, George Orwell

In the 1940s, when 1984 was written, television and television programming were only beginning to enter people's lives. Even with nothing but black-and-white sets to go on, George Orwell invented the telescreen, a device that could broadcast propaganda and watch you at the same time.

What he did not foresee was that when the real year 1984 arrived, it would be the internet that became the telescreen everywhere.

Leave aside government surveillance, which the law restricts in most countries. We depend more and more on tech companies, handing them our own information with both hands, and they have in effect become all-knowing and all-powerful.

Silicon Valley's companies are trying to resist the government's lengthening reach. In the 2016 standoff between the FBI and Apple, Apple refused in the end to build a backdoor into the iPhone, and led Facebook, Google, Microsoft and other tech companies in a joint defense of user privacy.

But nothing in business lasts. Yahoo was once a Silicon Valley pioneer too, the most valuable internet company in the world. It invented the idea of the web portal and changed the world of online information enormously.

In the second half of last year, Yahoo was hit by two large privacy scandals. In the first, acting on FBI instructions, it had spent a long time monitoring the data of hundreds of millions of email users, searching supposedly private mail by keyword.

If that could still be filed under government surveillance — something many people shrug at, having done nothing wrong — the second was plain disregard for users' digital security in the name of commercial interest.

Yahoo's user system had been breached and the email passwords of hundreds of millions of users leaked. To protect the company's image and avoid damaging a huge acquisition then under negotiation, Yahoo never disclosed the breach and never told users to change their passwords.

The leak exposed more than the accounts of hundreds of millions of email users. It could compromise their data elsewhere too, because many people use the same email-and-password combination on every internet service they touch.

All of this happened while Marissa Mayer was Yahoo's CEO. Before she moved to Yahoo, Mayer was Google's twentieth employee and spent 13 years there as a core executive. She was there when Google wrote "Don't be evil" into its IPO filing.

Yahoo sold off its core business in 2016 and changed its name this year. What it was is gone.

Like Yahoo, no one can say what big companies such as Apple, Google and Microsoft will be in ten or twenty years — who will run them by then, or whether the data they have collected will leak.

Other companies, in other places, are harder still to predict.

In 1984, O'Brien, an Inner Party member, could switch off his telescreen and escape being watched. Today, no one escapes.

Cover image: Pexels, Person of Interest