For all the positive things said about the system, a cloud has hung over Tesla Autopilot since Joshua Brown was killed while behind the wheel of a Model S last year. An in-depth investigation by the National Transport Safety Board (NTSB) has shed light on how Autopilot functioned leading up to the accident, revealing Brown had his hands off the wheel for the majority of the trip.
According to an in-depth report from the NTSB, logs from the car show Autopilot actively controlled the steering, acceleration and braking for just over 37 minutes leading up to the accident. For the majority of that period, the computer monitoring the steering wheel was in a state called "Hands Required Not Detected" – meaning the driver should be holding the wheel but isn't.
Seven times, that same computer swapped into "Visual Warning" mode, where a message pops up telling the driver to take the wheel. Every time the warning popped up, Brown put his hands on the wheel for a few seconds, before the car returned to "Hands Required Not Detected" state. Brown was holding the steering wheel during just 25 seconds of the trip prior to the accident.
Autopilot is technically capable of operating without a driver holding the wheel in some conditions, but drivers are told the system is "an assist feature that requires you to keep your hands on the wheel at all times," and the car displays a message to "always keep your hands on the wheel. Be prepared to take control at any time." Since the accident, Tesla has tweaked Autopilot so it doesn't work when the driver takes his or her hands off the wheel for an extended period.
This information doesn't explain why Brown and the Autopilot radar both failed to spot a tractor-trailer crossing the highway, and it wasn't designed to apportion blame for the crash. But it does prove drivers are mistaking Autopilot for a Level Five autonomous system, when it's actually closer to Level Two. Regardless of what the company calls it, the systems in Tesla cars are more of a semi-autonomous helper for now.