For years, the central question hanging over robotaxis has been simple: are they actually safer than human drivers? A new independent study from the Insurance Institute for Highway Safety (IIHS) offers some of the strongest evidence yet that they can be. Analyzing 50 million miles of Waymo's driverless operation, the IIHS found the company's robotaxis were involved in 68% fewer police-reportable crashes per mile than human drivers in the same cities — a landmark data point in the long, contested campaign to earn public trust in autonomous vehicles.
What the Data Shows
The IIHS examined 50 million miles of Waymo driving between 2021 and 2024 across San Francisco, Los Angeles, Phoenix and Austin, comparing it against 222 billion miles logged by human drivers over the same period in the same cities. The gap was substantial and consistent across categories:
- 68% lower overall police-reportable crash rate.
- 85% fewer single-vehicle crashes.
- 81% fewer crashes resulting in injuries.
- 91% lower rate of rear-ending another vehicle, and 40% lower rate of being rear-ended.
In raw terms, Waymo vehicles were involved in 1.28 reportable crashes per million miles, versus 4.06 for human drivers. The advantage varied by market — from 76% fewer crashes in Phoenix to 35% fewer in San Francisco — reflecting how local traffic, weather and road design shape autonomous performance.
The Caveats Matter
The researchers were careful to bound their claims. The findings are specific to Waymo's current Level 4 technology and the places it operated during the study; IIHS statistical services director Eric Teoh cautioned that moving into new cities with different conditions could change the picture. The team also had to discard roughly 25% of reported AV crashes because they were duplicates, occurred off public roads, or were not true crashes — a reminder of how messy the underlying data remains.
Crucially, the study does not vindicate driver-assistance features sold in consumer cars. Teoh stressed the results apply only to fully autonomous systems like Waymo's, not the partial-automation systems many drivers mistake for self-driving. And Tesla, whose Austin robotaxi began operating in 2025 with safety drivers and has logged comparatively few unsupervised miles, is absent from the 2021–2024 dataset entirely.
A Broken Data System
Perhaps the report's most consequential argument is not about Waymo at all, but about the inadequacy of the nation's AV safety data. The US requires companies to report incidents involving automated driving systems, yet most do not disclose how many miles their vehicles travel — and without that exposure data, crash rates cannot be calculated or compared across companies. The only reason this study was possible is that Waymo voluntarily publishes its driverless mileage.
IIHS urged reform, recommending that federal regulators require all AV companies to disclose vehicle miles traveled so meaningful, apples-to-apples safety comparisons can be made. Waymo endorsed the call, saying the results confirm its own peer-reviewed analyses and that "more high-quality data is fundamental for building public trust."
Why It Matters
Independent validation is a turning point for a technology that has run largely on company-supplied statistics. When a respected safety institute — funded by insurers, who have every incentive to price risk accurately — concludes that a robotaxi fleet crashes at a fraction of the human rate, the debate shifts from whether autonomous vehicles can be safer to how to scale that safety responsibly.
The commercial stakes are enormous. Robotaxi operators are racing to expand across dozens of cities, and credible safety evidence is the currency that unlocks regulatory approval, insurance, and rider confidence. Yet the study's data critique is a warning shot: as more operators deploy, the absence of standardized mileage reporting will make it increasingly hard to distinguish genuinely safe systems from the merely under-scrutinized. The next milestone the industry needs is not another impressive fleet — it is a data regime rigorous enough to prove which fleets deserve the public's trust.
