Self-driving cars have the potential to reduce traffic accidents by removing risks such as distracted, impaired, and fatigued driving. Early real-world evidence is promising: a 2026 Insurance Institute for Highway Safety study found that Waymo’s driverless vehicles were involved in 68% fewer police-reportable crashes per mile than human drivers in comparable locations.
However, those results apply to a specific Level 4 automated driving system operating in defined areas. They do not prove that every autonomous vehicle is safer than a human driver in every situation, and current technology cannot eliminate all crashes.
How Could Self-Driving Cars Reduce Accidents?
Fully automated vehicles could potentially prevent some of the mistakes that contribute to crashes when people are driving.
1. Eliminating distracted and impaired driving. An automated driving system does not text, become intoxicated, fall asleep, or lose concentration in the same way a human driver can.
2. Monitoring the road continuously. Autonomous vehicles may combine cameras, radar, lidar, and other sensors with software to monitor vehicles, pedestrians, cyclists, lane markings, traffic signals, and surrounding objects.
3. Responding to potential hazards. The system can reduce speed, brake, or change its path when it detects a developing conflict. This could help avoid some crashes caused by delayed human reactions.
4. Maintaining consistent driving behavior. Automated systems can be programmed to maintain following distances, follow traffic rules, and avoid emotional behaviors such as aggressive driving.
These advantages do not mean autonomous systems recognize every hazard perfectly or always react faster than a person. Their safety depends on the design of the system and the conditions in which it operates.
Are Self-Driving Cars Safer Than Human Drivers?
The strongest recent evidence comes from an IIHS study published on July 23, 2026. Researchers compared Waymo’s fully driverless operations with human driving in San Francisco, Phoenix, Los Angeles, and Austin.
| Crash measure | Waymo compared with human drivers |
|---|---|
| Police-reportable crashes | 68% fewer per mile |
| Injury crashes | 81% fewer per mile |
| Single-vehicle crashes | 85% fewer per mile |
The comparison was based on crash rates per vehicle mile traveled, not the raw number of crashes. Waymo accumulated about 50 million miles of driverless operation during the study period, while researchers compared those miles with human-driver data from the same cities and periods.
The study was published in 2026 but analyzed automated-vehicle crash data from 2021 through 2024. Researchers also adjusted the records to focus on incidents similar to crashes that human drivers would normally report to police.
Results differed by city. Waymo’s police-reportable crash rate was 76% lower than human drivers in Phoenix, 35% lower in San Francisco, and 71% lower in Los Angeles. In Austin it was 4% higher, although IIHS noted that the Austin sample was relatively small.
That distinction matters. The study provides evidence that Waymo’s Level 4 system performed better than comparable human driving in the conditions studied. It does not establish that every self-driving system, manufacturer, city, or future use case will produce the same safety improvement.
Why Can Self-Driving Cars Still Crash?
Autonomous driving does not remove every source of risk from the road.
Unusual road situations. Unexpected behavior by pedestrians, cyclists, other drivers, emergency vehicles, or unusual objects may create situations that are difficult for automated systems to interpret.
Bad weather and visibility. Heavy rain, snow, fog, glare, or other conditions can make it harder for some cameras and sensors to obtain useful information.
Road construction. Temporary lanes, hand signals from workers, unusual detours, and changing road markings can create situations that differ from normal traffic patterns.
Software and sensor limitations. Incorrect object classification, software errors, sensor failures, or incomplete information about the surrounding environment can contribute to unsafe situations.
Other road users. Even a vehicle that follows its programming correctly still shares the road with people who may run red lights, speed, suddenly change lanes, or behave unpredictably.
None of these situations automatically causes a crash, but they illustrate why removing the human driver does not remove all possible accident scenarios.
What’s the Difference Between Self-Driving Cars and Driver Assistance?
One of the most important distinctions in discussions about autonomous vehicle safety is the difference between systems that assist a driver and systems that actually perform the driving task without constant human supervision.
| Technology | Driver responsibility |
|---|---|
| Level 2 — Partial automation | The system assists with steering and acceleration/braking, but the driver continuously monitors the road and remains responsible |
| Level 3 — Conditional automation | The system performs the driving task under certain conditions, but the driver must be available to take over when requested |
| Level 4 — High automation | The system performs the driving task without human supervision within defined operating conditions or service areas |
| Level 5 — Full automation | The system could perform all driving under all road and operating conditions without a human driver |
Waymo’s driverless robotaxis operate as Level 4 vehicles within specified areas and conditions. Level 5 vehicles capable of autonomous operation under all conditions are not currently in widespread real-world operation.
By contrast, systems such as Tesla Autopilot and Full Self-Driving (Supervised), along with many other consumer driver-assistance products, are Level 2 systems that still require continuous driver supervision. They should not be treated as equivalent to the Waymo vehicles studied by IIHS.
That distinction is especially important because IIHS has found little convincing evidence that Level 2 partial automation itself reduces crash rates beyond the benefits already provided by crash-avoidance technologies such as automatic emergency braking and lane departure prevention.
Will Self-Driving Cars Eventually Eliminate Traffic Accidents?
There is currently no evidence that self-driving cars will eliminate traffic accidents completely. They may reduce certain categories of crashes, especially those caused by distraction, impairment, fatigue, or some types of human error, but other risks remain.
Vehicles will continue to encounter unpredictable road users, severe weather, unusual construction zones, equipment failures, and situations outside the conditions their automated systems were designed to handle.
There is also a major difference between demonstrating good safety performance in selected cities and deploying autonomous vehicles across the entire United States. Roads, weather, traffic patterns, infrastructure, and operating conditions vary enormously.
The current evidence therefore supports a more limited conclusion: some Level 4 self-driving systems can achieve lower crash rates than human drivers in the environments where they have been tested. How much autonomous vehicles ultimately reduce accidents nationwide will depend on the reliability of the technology, the conditions in which it operates, and how widely genuinely driverless systems are deployed.