Are self-driving cars safe?

Overall, yes, self-driving cars seem to be safe, but there are a few important caveats to keep in mind when comparing autonomous systems to human drivers.

First, let’s look at the available statistics. Google’s Waymo autonomous vehicles boast some of the field’s best technology, and they’ve logged over 10 million miles of real-world experience since 2009; as of 2018 (before they’d reached the 10-million-mile mark), they’d been involved in at least 30 minor accidents, but they’d only caused one—a lane-change incident that didn’t cause any injuries.

Proponents of autonomous tech often point to Waymo’s excellent track record to show the safety of self-driving cars. Skeptics and top industry voices, though, emphasize that it’s the context of the miles, not the number of them, that matters most.

“Miles traveled standing alone is not a particularly insightful measure if you don't understand what the context of those miles were,” Noah Zych, the head of system safety at the Uber Advanced Technologies Group, told Wired. “You need to know, ‘What situations was the vehicle encountering? What were the situations that the vehicle was expected to be able to handle? What was the objective of the testing in those areas? Was it to collect data? Was it to prove that the system was able to handle those scenarios? Or was it to just run a number up?’”

Google produces regular safety reports (link opens a PDF) explaining how its cars function in different environments, and Waymo subscribes to the quality-over-quantity theory, but the point stands: Even when equipped with state-of-the-art technology, self-driving cars won’t always respond to unexpected events in the same way a human driver would. Unconventional scenarios might prompt vehicles to make bad decisions.

In March 2018, a self-driving Uber vehicle struck a pedestrian in Tempe, Arizona, resulting in the first fatality attributed to an autonomous vehicle. According to a report from The Information, the vehicle detected the pedestrian but decided to stay the course, believing that the pedestrian was an inanimate object. Tempe police determined that the vehicle’s human operator could have avoided the accident, were she not streaming videos on her phone at the time of the collision.

These types of incidents don’t necessarily mean self-driving vehicles are unsafe, especially when compared with human drivers.

“Of all serious motor vehicle crashes, 94 percent are due to human error or choices,” the National Highway Traffic Safety Administration (NHTSA) says on its website. “Fully automated vehicles that can see more and act faster than human drivers could greatly reduce errors, the resulting crashes, and their toll.”

However, NHTSA also notes that true self-driving vehicles aren’t on the roads—yet.

“There is no vehicle currently available for sale that is ‘self-driving,’” the administration writes. “Every vehicle currently for sale in the United States requires the full attention of the driver at all times for safe operation.”

With current autonomous vehicles, a human driver needs to remain engaged with the vehicle and be prepared to take control at a moment’s notice. As that changes in the coming decades, drivers will likely benefit: According to a report (PDF) from KPMG’s insurance task force, by 2050, the adoption of autonomous vehicle technology could reduce the frequency of accidents by almost 90 percent.

For that to happen, legislators will need to determine fault for accidents caused by driverless cars, and that’s no small task. Still, the benefits seem clear—autonomous cars will come with a unique set of safety concerns, but they’re likely much safer than human-driven vehicles.