Tesla has published its first quarterly vehicle safety report, which CEO Elon Musk promised in May would show “exactly what Autopilot’s safety [level] is.” In it, the company says Tesla owners got into one “accident or crash-like event” every 1.92 million miles driven without Autopilot during the third quarter of this year. When drivers did use Autopilot, Tesla says, they experienced one accident or crash-like event every 3.34 million miles.
Tesla doesn’t include any more detail than that, though. It doesn’t describe any common traits between these crashes, or where and when they typically occur. The company also doesn’t offer any insight into how often its drivers disengage the system, which has become a go-to statistic in the autonomous vehicle industry. Other automakers like Ford and GM, as well as Waymo, have issued safety reports regarding their autonomous vehicle testing as part of a voluntary program started by the Department of Transportation.
Tesla reworked how it collects data about Autopilot usage, but didn’t say how
The safety report was also published on the same day that Consumer Reports put Autopilot second in its first-ever ranking of the best driver assistance systems currently available. GM’s Super Cruise feature came in first.
In outlining this report, Tesla says it designed an entirely new method of organizing and collecting data from customer vehicles. “This new data stream allows us to gather the most critical fleet-wide statistics from the exact moment a crash-related event is detected by our system,” the company wrote.
Citing the National Highway Traffic Safety Administration, Tesla says there is a car crash every 492,000 miles when looking at all vehicles. Comparing that to Tesla’s base crash statistic — one every 1.92 million miles driven — might give the impression that the company’s cars somehow are involved in far fewer accidents. But Tesla includes “accidents as well as near misses (what we are calling crash-like events)” in its calculation, which makes its comparison with government data questionable. The company did not respond to questions about how it calculates and identifies near misses or crash-like events.
Tesla has closely guarded statistics about Autopilot’s performance and safety in the past, but a succession of highly publicized crashes since 2016 — some fatal — raised questions about the driver assistance feature’s effectiveness.
Autopilot has been scrutinized for years
On a quarterly call with investors this past May, Musk responded to some of those questions. He chastised the media for its coverage of crashes where Autopilot was involved, saying that news outlets write “inflammatory headlines” about the feature that “are fundamentally misleading to the readers.”
Musk said this “negative news” had a material impact on Autopilot usage. “I was like, okay, this is not good, because people are reading things in the press that cause them to use Autopilot less, and then that makes it dangerous for our customers,” he said. The CEO promised during that same call that Tesla would start issuing quarterly reports about Autopilot’s safety performance beginning in the second quarter of the year.
One of the common criticisms levied against Tesla’s Autopilot, especially after a driver died in a crash in Mountain View, California earlier this year, was that the system does not do enough to ensure the person behind the wheel pays attention to the road. Tesla itself even said that the driver who died had received multiple warnings before his car crashed into a highway divider.
Super Cruise, which is Cadillac’s semi-automated driving system, is comparable to Autopilot in many respects. But it uses more advanced and aggressive driver monitoring, which is why Consumer Reports ranked it highest on its new list.
Tesla places sensors in the steering wheel to determine whether or not the driver has their hands on the wheel. If they don’t, the car will start a cycle of alarms that increase in pace and volume before Autopilot automatically disengages and the car pulls itself over to the side of the road.
Consumer Reports says Tesla’s driver monitoring is “insufficient”
Some critics have said that’s not enough, and Consumer Reports agrees. “This is an insufficient way of measuring driver attention, and it provides little assurance that the driver is even awake,” the publication writes. “Because of the impressive ability of Tesla’s Autopilot to keep the vehicle centered in its lane, it’s easy for drivers to become overreliant on it.”
Super Cruise relies on eye-tracking technology to make sure that the driver is paying attention to the road. If the system thinks that a driver’s gaze is wandering, or they’re falling asleep, a light on the steering wheel flashes red, the seat vibrates, and alerts sound. Consumer Reports says Super Cruise “does the best job of balancing high-tech capabilities with ensuring that the car is operated safely and that the driver is paying attention.”
Elon Musk said earlier this year that he rejected using eye-tracking technology in Tesla’s cars because it was “ineffective.”
Tesla says it will release safety reports every quarter going forward, and that it also plans to start “gathering serious injury data from [the company’s] customers following an accident.” The company didn’t offer any guidance as to whether that data, or other statistics about Autopilot, will be included in future reports.