Self-driving and semi-autonomous vehicles have changed the conversation about road safety, technology, and who is responsible when something goes wrong. Cars that were once confined to labs and closed test tracks now operate on city streets across the United States and around the world. These vehicles are expected to reduce crashes caused by human error, but real-world crash data is essential to understanding how they actually perform and what risks they carry.
Drawing on trusted sources such as the National Highway Traffic Safety Administration (NHTSA) and its Standing General Order on crash reporting, this article looks at self-driving car accident statistics from 2019 through 2024 - crash trends, company-by-company data, common causes, and key safety takeaways. All of it underscores how important legal guidance is after a self-driving car accident.
What counts as a self-driving car accident?
Before looking at the numbers, terminology matters. The media often uses the term self-driving loosely, but safety reports rely on precise definitions tied to the vehicle's level of automation. Automated Driving Systems (ADS) refer to higher-level automation - Levels 3 through 5 - where the system can handle the full driving task under certain conditions. Advanced driver assistance systems (ADAS), at Levels 1 and 2, provide features like lane keeping and adaptive cruise control but still require a human behind the wheel.
Reporting requirements and data limitations
NHTSA's Standing General Order requires manufacturers and operators to report certain crashes involving ADS and Level 2 ADAS within 24 hours. Even so, the data has real limitations, and the figures below should be read with that in mind.
How many self-driving car crashes happen each year?
NHTSA began collecting nationwide ADAS and automated-vehicle crash data in June 2021, with some state and company reporting going back to 2019. Reported crashes have climbed steadily since then - driven by wider adoption of driver-assistance systems, expanded testing areas, and better reporting compliance, not necessarily by more dangerous technology.
Crashes by company
Crash counts often reflect both the size of a company's fleet and how aggressively it has rolled out automation:
- Tesla (Level 2 ADAS): more than 2,000 reported crashes, reflecting a large nationwide fleet; many driver-assistance crashes involve inattentive drivers.
- Honda (Level 2 ADAS): more than 100, mostly involving lane keeping and adaptive cruise control.
- Subaru (Level 2 ADAS): more than 50, tied to the widely deployed EyeSight driver-assistance system.
- Waymo (Level 4 ADS): more than 1,000, from fully driverless operations in Phoenix and parts of San Francisco.
- Cruise (Level 4 ADS): more than 150, including notable incidents in dense urban settings like San Francisco.
- Other automakers: roughly 50 combined, including Ford, Toyota, and Mercedes-Benz test programs.
The key insight: a high crash count for a company like Tesla does not necessarily mean its technology is more dangerous. Large fleets with nationwide coverage naturally generate more reports, while the numbers for an ADS company like Waymo reflect a much smaller operating footprint.
Where do these crashes happen?
Self-driving car crashes are concentrated in states with permissive testing laws, densely populated urban areas, and climates that are friendly to the vehicles' sensors. That is why a handful of states and cities account for most of the reported incidents.
Injuries and deaths
Crashes involving automated vehicles remain a small share of all U.S. traffic accidents, but severity is the safety indicator that matters most. Even with relatively few total crashes, these systems should be judged on how well they prevent the most dangerous collisions - not just on the raw numbers.
Are self-driving cars safer than human drivers?
The promise of the technology is reducing human error, which plays a role in more than 90 percent of traffic accidents. So far, though, research comparing self-driving cars with human drivers has not reached a clear conclusion. The evidence suggests automated vehicles can prevent many crashes caused by lapses in attention and judgment - but only when they are designed to put safety ahead of speed and efficiency.
Public perception
Despite the technical progress, surveys by industry and government point to mixed public trust in self-driving cars. Perceptions often shift after high-profile incidents, so transparency and a consistent safety record will be critical to building confidence.
Who is liable in a self-driving car accident?
Sorting out fault in a self-driving car accident can be far more complicated than in an ordinary crash. Depending on the facts, responsibility may rest with the driver, the vehicle's manufacturer, or the company operating the system. If you have been involved in a crash with an automated vehicle, talk with an experienced car accident attorney about your legal options.
Frequently asked questions
What is the difference between ADS and ADAS?
ADS refers to higher-level automation (Levels 3-5) that can perform the full driving task under certain conditions. ADAS (Levels 1-2) provides driver-assistance features such as lane keeping or adaptive cruise control but still requires human supervision.
How do authorities investigate self-driving car crashes?
NHTSA collects crash reports from manufacturers, and independent state or federal investigations often follow. Data from the vehicle's sensors, cameras, and system logs serves as key evidence.
Can I sue after being hit by a self-driving car?
Yes. Depending on who was negligent, you may have a claim against the driver, the manufacturer, or the operator. Legal advice is essential to pin down liability in these cases.
Are fully autonomous vehicles on public roads today?
Yes, but only in limited geographic areas. Companies such as Waymo and Cruise operate Level 4 vehicles in a handful of U.S. cities.
The risks and the reality
Self-driving technology brings both promise and challenges. The data shows automated vehicles can reduce some of the risks human drivers create, but they are far from accident-free. Early crash statistics deserve careful interpretation, because fleet sizes, reporting standards, and deployment areas vary widely from place to place. As testing and deployment expand, ongoing monitoring, transparency, and informed discussion will be essential to understanding - and improving - how safe these vehicles really are.
This article is general information, not legal advice about your specific situation.