The development of autonomous vehicles (AVs) by companies like Waymo, which originated from Google’s self-driving project, represents a fundamental shift in transportation safety. This technology promises to remove human error, which is the dominant factor in most collisions. Public acceptance and regulatory approval depend on transparently demonstrating safety performance using verifiable metrics. Analyzing AV safety requires looking beyond simple incident counts to understand data reporting methods, collision frequency, and resulting injury risks for occupants and other road users. This article examines the safety data Waymo has published, focusing on these dimensions to provide a clear picture of the technology’s current accident profile.
Methodology for Measuring Safety Performance
Quantifying the safety of an autonomous vehicle system demands standardized and consistent reporting metrics. The primary measure for calculating accident rates is “miles driven,” specifically “Rider-Only” (RO) miles, which represent the distance covered without a human safety driver present. Through June 2025, Waymo reported accumulating over 96 million RO miles, providing a substantial exposure base for statistical analysis. This mileage serves as the denominator for calculating incidents per million miles (IPMM), allowing for direct comparison against human driving statistics.
The definition of a “reportable accident” for AVs is significantly broader than for conventional human-driven vehicles. Autonomous vehicle operators must report any physical contact resulting in property damage, injury, or fatality under the National Highway Traffic Safety Administration’s (NHTSA) Standing General Order (SGO). This strict standard means AV data includes minor scrapes that human drivers often do not report to the police or insurance companies, inflating the raw incident count relative to the human benchmark data. Researchers must therefore apply statistical adjustments to human crash data to account for this significant underreporting bias.
Another metric used to gauge system maturity is the “disengagement report,” which tracks instances when the AV system yields control to a human safety driver or operator. A disengagement is not an accident but rather an intervention prompted by the vehicle’s inability to safely navigate a scenario or by the safety driver perceiving a risk. In 2023, Waymo vehicles recorded a disengagement roughly every 17,311 miles of testing, demonstrating the system’s increasing reliability. Analyzing the causes of these disengagements provides insight into the limitations of the software, such as complex unprotected turns, which helps target future system improvements.
Analysis of Accident Frequency and Cause
The frequency of collisions compared to human drivers operating in the same conditions is the primary measure of safety performance. Waymo’s autonomous vehicles demonstrate a significantly lower accident rate across various severity metrics when benchmarked against human drivers in the same cities. For all police-reported crashes, the Waymo Driver had a rate of 2.1 incidents per million miles (IPMM), which represents a 57% reduction compared to the human benchmark of 4.85 IPMM. The reduction is even more pronounced when considering crashes involving any injury, where the Waymo rate of 0.41 IPMM is an 85% reduction against the human rate of 2.78 IPMM.
Analysis of the fault and cause in Waymo-involved collisions reveals that the overwhelming majority of incidents are caused by human error from the other party. In a review of serious crashes, the Waymo software was determined to be at fault in only a small handful of cases. This pattern underscores that the Waymo Driver’s most challenging safety scenario is often reacting to unpredictable, unlawful, or inattentive human behavior.
The most common types of collisions involving Waymo vehicles are often those where the AV is a passive participant. Waymo’s software is programmed for defensive driving, which includes maintaining larger following distances and braking early. This sometimes leads to rear-end collisions from frustrated or inattentive human drivers. Furthermore, the system has shown a statistically significant reduction in specific crash types compared to human drivers. Injury-causing intersection accidents were reduced by 96%, and side-impact crashes saw a 74% reduction, suggesting the AV excels at complex maneuvers and cross-traffic awareness.
The comparison to human driving statistics highlights that while AVs are not accident-free, their risk profile is fundamentally different. Human drivers are responsible for an estimated 94% of crashes, often due to speeding, distraction, or impairment. Waymo vehicles adhere perfectly to speed limits and traffic laws, resulting in a driving style that significantly reduces high-severity, high-speed incidents.
Comparison of Occupant Injury Outcomes
The ultimate measure of safety is the outcome for the people involved in a collision, and the data suggests Waymo crashes result in significantly lower injury severity than those involving human drivers. The autonomous system’s defensive driving style inherently reduces the energy of a collision, as the AV will detect a potential hazard earlier and attempt to brake or maneuver defensively. This behavior helps explain the reported 81% fewer crashes with deployed airbags compared to human drivers, indicating a substantial reduction in high-impact forces.
Waymo vehicles are also involved in 85% fewer crashes resulting in serious injuries when compared to the human benchmark. While the total number of incidents reported to NHTSA between 2021 and 2024 included 47 total injuries, only three were categorized as serious, with the remaining being minor or moderate. Even in the rare instances of a fatality or serious injury, those cases were overwhelmingly caused by the external human-driven vehicle, not the Waymo system.
Beyond driving behavior, the vehicle platform itself incorporates design considerations aimed at occupant protection in an AV context. The cars are equipped with redundant safety systems, and the overall design is part of a broader strategy aimed at achieving a goal of zero preventable fatalities. The focus on lower-severity crashes is a direct result of the AV software prioritizing caution over typical human driving aggression.
The significant reduction in injury claims, including an 88% reduction in claims for serious injury or death compared to human drivers, underscores the safety advantage. This is particularly evident in vulnerable road user protection. Waymo reported a 92% reduction in injury-causing crashes involving pedestrians and an 82% reduction for cyclists. The Waymo Driver’s superior 360-degree awareness and consistent, cautious behavior appear to be highly effective at mitigating the most severe risks to all road users.