A new study from the Insurance Institute for Highway Safety gives the strongest third party support yet for the claim that driverless cars crash less than people do. On the InsideEVs podcast, the hosts run through the numbers: across San Francisco, Phoenix, Los Angeles and Austin, Waymo vehicles were involved in 68 percent fewer crashes per mile than human drivers, along with 85 percent fewer single vehicle crashes and 81 percent fewer injury crashes. The data covers 2021 to 2024, roughly 50 million driverless miles against about 222 billion human miles in those same cities. Then the episode does something more useful than repeating the headline, and explains why one of the study's own authors called producing it unsustainable.
The problem is that the two sides of the comparison are not counted the same way. The podcast describes the federal reporting rule for automated vehicles as covering essentially any impact involving property damage, injury or death, while human crashes reach a database only if someone calls the police or files a claim. The podcast cites one of the study's authors saying that when the reported automated incidents were checked against police reported human data, roughly 80 percent of them were the kind of thing a normal driver would never have reported. Across 736 public road crashes with automation engaged, the podcast says only about 22 percent were estimated to be police reportable. There is a second asymmetry the episode raises: driverless fleets avoid the messiest environments, and comparing a curated operating area against everything humans drive through is not a like for like test. Waymo also discloses its mileage voluntarily, which most operators do not.
The rest of the episode is worth the time. On the Ferrari Luce, the hosts say early prototype drives from enthusiast outlets have been strongly positive despite the styling backlash, and run through the specs as reported: roughly $640,000, five seats, a 122 kWh pack, an 800 volt architecture and 0 to 60 in about 2.5 seconds. They cite Financial Times reporting that Ferrari has already hit a 2026 target of 500 units, while questioning how meaningful a target disclosed after the fact really is. The Rivian R2 section is more sobering. Testing by Out of Spec, cited in the episode, has the R2 using between roughly 18 and 27 percent more energy than a Model Y at steady speeds, covering 255 miles at 70 mph against a 330 mile EPA rating, while charging 10 to 60 percent in 15 minutes with a peak of 229 kW. The hosts' conclusion is that the EPA cycle, weighted toward city driving, no longer describes how people actually use these cars, and they argue the rating should be a 70 mph highway figure instead. Two other items get airtime. New J.D. Power data cited in the episode puts electric vehicle buyers at 910 on a 1,000 point satisfaction scale against 850 for everything else, with powertrain feel scoring particularly well. And the hosts flag a claim from Chinese manufacturer FAW of a battery that charges 10 to 90 percent in eight minutes, noting that nobody yet knows what that pace does to long term battery life.
Bottom line: The autonomy debate keeps getting fought with numbers that were never built to be compared, and this episode is the clearest explanation of that problem to come along in a while. Waymo almost certainly is safer than the average driver. Nobody can currently prove it to the standard a regulator should demand, and the company doing the most voluntary disclosure gets punished for it every time a rival stays quiet. A national reporting standard would fix this in a year. Until then, treat every safety percentage from any operator, including the flattering ones, as a starting point rather than a finding.
Commentary on a third-party video. Figures and claims are as presented in the source and have not been independently verified. Spotted an error? Tell us and we will correct it.