Waymo says its robotaxis prevented 841 injuries over 270 million miles

Waymo says its driverless vehicles prevented 841 people from being injured in traffic crashes across 270 million autonomous miles driven through June 30, 2026. The company published the figure in a September 24 safety-data release, framing it as evidence that removing a human driver from behind the wheel measurably cuts the number of people hurt on the road.

The claim rests entirely on Waymo’s own comparison against human-driver crash rates in the same five metro areas where its robotaxis operate. Independent researchers who have examined a narrower slice of the same data arrive at a real but noticeably smaller number.

Counting 841 Avoided Injuries Across Five Cities

Waymo’s rider-only service now spans Los Angeles, the San Francisco Bay Area, Phoenix, Austin and Atlanta, and the company says the 270 million miles logged there produced 82 percent fewer injury-causing crashes than a human-driven benchmark would predict, a reduction it translates into 841 avoided injuries. Of those, the company says 175 would have involved a pedestrian, cyclist or motorcyclist, and 55 would have been serious injuries or worse.

Serious-injury-or-worse crashes fell even further in Waymo’s accounting: a 95 percent reduction, or roughly 20 times fewer than the human-driver comparison. Pedestrian injury crashes dropped 93 percent, cyclist injury crashes 86 percent, and motorcyclist injury crashes 82 percent, according to the same release — the category of victim least able to protect themselves in any crash showing the largest claimed improvement.

An 82 Percent Drop, Built on Police Data and a Statistical Adjustment

The numbers are not self-certified in a vacuum. Waymo’s own safety-impact methodology page says the comparison draws on state police-reported crash records and vehicle-miles-traveled data from the counties where it operates, then applies what it calls a dynamic benchmark that adjusts for where within those counties its vehicles actually drive. Waymo’s crash data itself flows from a federal source: the National Highway Traffic Safety Administration’s Standing General Order requires any company operating an automated driving system to report a crash whenever the system was active within 30 seconds of impact and the incident caused injury, towing, or hospitalization.

The requirement has been in place since 2021 and was last amended in April 2025, and it is the same mechanism that generates monthly public crash data on every robotaxi operator, not just Waymo, giving outside researchers a way to check company-reported figures against a common federal dataset rather than taking a safety release at face value. Even so, 270 million miles is a small sample against U.S. driving as a whole: Electrek’s review of the release noted that human drivers log roughly 3 trillion miles a year nationwide, and injury crashes are rare enough events that a company’s own multi-year mileage total still represents a sliver of the driving population any safety claim is being measured against.

Independent Research Lands on a More Conservative Number

An outside version of that check has already happened once this year. The Insurance Institute for Highway Safety published its own analysis in July 2026, examining roughly 50 million driverless Waymo miles against 222 billion human-driven miles logged in the same cities between 2021 and 2024. Researchers there worked from 736 public-road crash reports, discarded about a quarter as not meeting a police-reportable threshold, and were left with 89 qualifying Waymo crashes, 64 of them in fully driverless mode. The result: a 68 percent lower rate of police-reportable crashes overall — a meaningful safety edge, but well short of Waymo’s own 82 percent injury-crash figure and its 95 percent serious-injury figure.

IIHS President David Harkey said “the results show that, on a limited scale, these driverless cars are safer than human drivers — who can be impaired or drowsy or suffer lapses in attention.” His organization’s study also found the advantage varied sharply by city: crash rates 76 percent lower in Phoenix and 71 percent lower in Los Angeles, but only 35 percent lower in San Francisco, and 4 percent higher in Austin than the human baseline there.

Harkey’s research flagged a structural problem behind every one of these percentages, on either side of the gap: no federal requirement currently forces any automated-driving company to disclose its total miles driven or how many of its trips still involve a human safety operator, which means every crash-rate comparison, Waymo’s included, is built on data the company reporting the crashes also controls.

Vulnerable Road Users See the Sharpest Numbers on Both Counts

Where Waymo’s and IIHS’s numbers agree in direction, if not in size, is on who benefits most. Both analyses point to pedestrians, cyclists and motorcyclists as the group seeing the largest relative safety gains from removing a human driver — the population least protected by the vehicle itself and most dependent on a driver’s attention and reaction time in the moment before impact. Whether that gain is closer to Waymo’s claimed 93 percent or the more conservative range implied by IIHS’s citywide breakdown remains an open question the next round of federal crash data will help settle.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview