The Data Piles Up: Robotaxis and Driver-Assist Tech Are Crashing Far Less Than Humans

A growing body of research is undercutting public skepticism about self-driving cars. Even the mandated building blocks of autonomy are already paying off: the IIHS found that automatic emergency braking tuned to detect people cuts pedestrian crashes by 27 percent, while automated braking slashes rear-end collisions by half and their injuries by 56 percent. Insurance data tells the same story, with property-damage claims falling 13 percent for AEB alone and up to 39 percent for cars bundling adaptive cruise control and lane-departure warnings. These systems will be required on all U.S. light vehicles by 2029, and most new models already comply.

At full (Level 4) autonomy, Waymo now reports 20 million paid rides across 220 million miles, and both its own study and an independent July IIHS analysis point to dramatic safety gains—roughly 92 percent fewer serious-injury crashes in Waymo’s numbers, and 68 percent fewer crashes overall (81 percent fewer injury crashes per mile) in the IIHS comparison against human drivers in Phoenix, LA, San Francisco, and Austin. The caveats matter, though: robotaxis operate only in fair weather in a handful of cities, roughly half of human crashes go unreported (which flatters human stats), and there are still no national AV performance standards—just a patchwork of state rules that makes clean comparisons hard.

The regulatory landscape is shifting quickly. NHTSA granted Amazon’s Zoox the first-ever exemption from motor-vehicle safety rules for its steering-wheel-free robotaxi, and launched a three-year, $5 million consortium with SAE to build the country’s first AV competency standards. Some public-health researchers argue AVs should be framed not as a tech moonshot but as a life-saving intervention on the scale of seat belts—given that road crashes kill about 40,000 Americans and 1.16 million people worldwide each year. A skeptical note from the discussion: the U.S. traffic death rate (14.2 per 100,000) already dwarfs the UK’s 2.6 and the EU’s 6.7, suggesting proven policy and infrastructure fixes could save lives without waiting on autonomy.