Will self-driving cars make police chases a thing of the past?

THE BRAND-NEW Ford DeLuxe Fordor that Bonnie Parker and Clyde Barrow plucked from a Topeka, Kansas, driveway in 1934 was top-shelf: Cordoba gray paint, a big chrome grill, and a sumptuous leather interior. But what really attracted the car-savvy bank robbers lay beneath the hood—its 85-horsepower flathead V-8 engine. This new breed of machine was designed to give every American, opportunistic crooks included, a taste of speed. Every American except the cops. Depression-era police departments suffered from tight budgets and outdated tech; many officers still bumped around in old Model As, which had half the Fordor’s muscle. People like Parker and Barrow could, of course, steal any ride they wanted, and they chose wisely. It took a Texas Ranger with his own V-8, and a small-town cop who went through cars, trucks, and even a limousine, before the law finally caught up with—and killed—the outlaw couple.

That’s an old story, but 84 years after Parker’s and Barrow’s deaths, car chases remain common: In 2012 alone, police in the United States racked up 68,000 pursuits, according to the latest data from the Bureau of Justice Statistics. Half of these chases started with a simple traffic stop—a broken taillight, failure to signal. But then, for various reasons, these drivers presumably had reasons to rabbit, from outstanding warrants to smoldering joints to illegal weapons. Whatever the trigger, most police pursuits today are pretty short. Two-thirds end in fewer than 3 miles, but even a quick chase can turn deadly. USA Today noted that 11,506 people died between 1979 and 2013 as the result of car chases. Many of them were bystanders.

It’s because of these grim statistics that police brass now discourage high-speed pursuits. “We do traffic enforcement to make people safe,” says Lt. David Ferry of the Los Angeles Police Department. “If I’m going to chase someone because they ran a light, I’m going to create a more dangerous scenario than if I don’t take action.” So Ferry and his colleagues aim to be more deliberate. LAPD helicopters radio the direction of suspect cars to police cruisers below, allowing cops to make safer pursuit decisions. “We won’t be right up on you, but we’ll still get you,” Ferry says. Oftentimes, they’ll just call in a scofflaw’s plate and check for violations. If there are none, they’ll simply mail a summons. Yet the LAPD gave chase 749 times in 2017, up from 394 in 2014.

The state of our streets might seem bleak, but the coming age of autonomous vehicles should shift us into a safer gear. If law-abiding conveyances can move about without drivers sitting at 100-percent attention, and with them, many traffic violations—both accidental and calculated—reducing the need for patrol cars altogether. As self-driving and driver-assistance technologies replace our fallible senses with precision radars, and exchange our impulsive brains for careful computers, four-wheeled mischief will come to a complete stop.

Right?

Maybe, but the very trustworthiness of autonomous vehicles (AVs) will actually give bad guys an entirely new criminal toolkit, according to Todd Humphreys, an engineering professor at the University of Texas at Austin with a knack for cracking everything from cellphones to cars. “Some crime will be incredibly sophisticated, and some will be incredibly unsophisticated,” he says.

The most artless could be the deadliest. Thomas Cowper, a retired state trooper and former member of an FBI-sponsored group gaming out impending security threats, says AVs could make acts of terror scarily easy. Stash an explosive in a vehicle—be that a robo taxi or an airport shuttle—set its destination for a busy tourist spot, and boom: fiery hellscape, no hacking required. Police could establish bomb-sniffing-dog checkpoints in places like Times Square or at the Hollywood Walk of Fame, or they could run every car through supersize airport body scanners, but that undertaking would be enormous.

Things get even trickier when would-be criminals get their hands on a car’s schematics. Missy Cummings, a former US Navy fighter pilot and a professor in Duke University’s Pratt School of Engineering, says turning an autonomous vehicle into an agent of chaos will require only a basic understanding of how it works. Lidar, for example, is the laser-based detection system that sits atop most self-driving test cars today. You don’t have to know what it is to disrupt it, only where it is. Something as mundane as dust on its lens could blur the lidar’s perception of traffic, Cummings says.

Of course, police will also have new tools at their disposal. They could, according to Cummings, stop a sketchy self-driver by shooting a burst of silly string at its lidar module. “You wouldn’t even have to use bullets.”

Lidar is hardly the only vulnerability. The typical driverless toolkit consists of three other susceptible sensors: cameras at the front and back, GPS connected to positioning satellites, and a radar that bounces radio waves off nearby objects to let the car know its place in space. Artificial intelligence analyzes the data from all four sensors and uses it to inform future decisions. A software-savvy ne’er-do-well could do significant damage to any part of this self-driving suite, putting the vehicle’s occupants—and everyone else on the road—at risk.

A determined criminal, Humphreys says, could jailbreak the car’s central brain by altering the software and allowing the hacker to speed, or direct the car manually. A culprit will be able to manipulate other vehicles too. Vehicle-to-vehicle communication systems may also let AVs cruise, merge into traffic, and exit highways in sync with other cars on the road. By sending false messages, Humphreys says, you could clear freeways, cause crashes, and even turn police roadblocks against themselves.

Hackers love to mess with GPS too. In 2016, a journalist in Moscow discovered his car’s GPS indicated it was 20 miles away, at the city’s airport, when it was actually cruising through the heart of downtown. The next summer, Russian hackers appear to have put similar GPS-distorting technology to work in the Black Sea; attackers spoofed at least 20 ships, disrupting each captain’s ability to determine the distance between their ship and others in the highly trafficked water way. Using the same tools, Humphreys says, you could send vehicles and their occupants veering into oncoming traffic or careening off a cliff.

Still, a well-planned hack doesn’t guarantee you’ll get away—at least not for long. That’s because autonomy doesn’t just replace drivers; it will probably replace the entire car-purchasing experience. Automakers may deploy fleets of AVs as leased rather than owned vehicles, possibly using a subscription model. When a company’s intellectual property is in the driver seat, it’s in the firm’s financial interest to monitor all of its vehicles all the time—and shut down a rider if it detects tampering.

Tesla has already flexed this ability, rolling out software updates to its vehicles over the air. When the company wanted to adjust its cars’ brakes, it deployed code that reduced the stopping distance in its rides by 19 feet. Ad hoc fixes are also available, as the company demonstrated recently when it had to push a fix to address a glitch that pinched fingers in windows.

It’s that kind of vast connectivity that will prove the biggest deterrent to people fleeing in, or misusing, their robo-chauffeurs. Computer-chipmaker Intel predicts that autonomous vehicles will generate 4 terabytes of data for every 90 minutes of travel, enmeshing us all in a sort of self-policing police state. Couple that with the location and license-plate information that our current crop of traffic and toll-bridge cams already collect, and no matter how far you get, you could still end up in cuffs. 

It might not even take a lot of techno-sleuthing to bring you in. Ask most any cop about their long-term hopes, and they’ll eventually settle on two words: kill switch. Instead of speeding down the highway after crooks, many await the day automakers create a simple button capable of shutting down riders instantly—and remotely.

This might sound like a dystopian fantasy, but these science-fiction-style tools already exist. When requested, General Motors’ subscription security service OnStar can use GPS to locate a stolen vehicle, and even slow it to a snail-like 3 miles per hour.

That doesn’t mean cops take kindly to every new invention. Ford’s patent for a car that can autonomously chase vehicles and remotely issue citations isn’t exactly great for job security. Electromagnetic pulses that can disrupt your car’s computers are already available, but not in the neighborhood precinct. Units like the LAPD “are pretty scared about EM pulses and directed energy devices,” Ferry says.

While academics, engineers, and law-enforcement officials look to the future, delinquents would be wise to hit the history books—or at least a used-car lot.

Even if all manual vehicles are outlawed (a scenario most experts find unlikely), or simply barred from autonomous-only roads (more probable), black-market deals and nostalgic gearheads will ensure that human-piloted hot rods persist. Drive one, and you’ll get your way on the streets. “If you go rogue in a legacy vehicle,” Humphreys says, “other vehicles are likely to give you a wide berth.” That’s because self-driving cars will prioritize human life. In a sea of Miss Manners machinery, you’ll be able to speed, blow through stop signs, and clear a path to freedom, exploiting other vehicles’ bias for politeness and lawfulness, Humphreys says.

The car chase of the future, it seems, isn’t all that futuristic. It might rely on the same manual controls from chases of yore. Whether it’s a white Bronco, or even a Ford Fordor DeLuxe, a modern-day Bonnie and Clyde will always find a way to slide into the 6 o’clock news.

This story originally ran in the Winter 2018 issue of Popular Science. Read more PopSci+ stories.