How safe are autonomous vehicles? 2024

Author pictureAuthor picture
Author picture
Author picture
Edited by:
self-driving car on a highway with passengers

Autonomous vehicles are equipped with systems that monitor and control their operation. They employ a variety of technologies to respond to their environment, including GPS, infrared sensors, cameras, LIDAR, RADAR, which are wireless communications with other vehicles or infrastructure, such as stoplights. While fully automated vehicles are not available to the public, they are being tested on some public roads. Less advanced automated driving systems are already installed on many modern automobiles. With millions of traffic crashes resulting in tens of thousands of fatalities in the United States every year, some are hopeful that autonomous vehicles can provide a safer alternative to human drivers.

Key insights

In 2021, there were 6.1 million traffic crashes in the U.S., resulting in 42,939 fatalities, with many linked to human error.

Jump to insight

Between July 2021 and December 15, 2023, there were 508 reported autonomous vehicle crashes in the U.S.

Jump to insight

Between July 2021 and December 15, 2023, data shows 1,239 crashes involving a Level 2 automation system, such as Tesla’s Autopilot, in the U.S.

Jump to insight

A 2023 study forecasted that Advanced Driver-Assistance Systems (ADAS) could prevent 37 million traffic crashes in the U.S. between 2021 and 2050, averting 14 million injuries and almost 250,000 fatalities.

Jump to insight

Autonomous vehicle levels

The Society of Automotive Engineers (SAE) has established six Levels of Driving AutomationTM to classify vehicle automation systems rather than the vehicles themselves. The lowest, Level 0, is defined as No Driving Automation, while Level 5 is categorized as Full Driving Automation.

These levels can be grouped into two categories: Levels 0 through 2 are each considered an Advanced Driver-Assistance System (ADAS), while Levels 3 through 5 are each classified as an Automated Driving System (ADS). Currently, consumers can only purchase vehicles equipped with ADASs, while vehicles carrying ADSs are highly regulated and can only be deployed on public streets for research purposes.

Society of Automotive Engineers (SAE) Levels of Driving AutomationTM explained

Benefits of autonomous vehicles

  • Fully automated safety features reduce crashes and prevent injuries
  • Provide transportation access to people with disabilities and underserved communities
  • Potentially reduce greenhouse emissions and decrease environmental costs

While ADS-equipped vehicles are still a very nascent technology, there are a number of ways they could someday benefit society.

The National Highway Traffic Safety Administration (NHTSA) has identified the Five Eras of Safety for automobiles, beginning in the 1950s with basic features, such as seatbelts. The NHTSA predicts the fifth era will begin in 2025 with the adoption of fully automated safety features. If they operate more safely than humans, fully automated systems could reduce crashes and prevent injuries, deaths, and potentially save billions of dollars in economic costs. The NHTSA estimates that, in 2019, 9% of the total cost of motor vehicle crashes in the U.S. was paid through tax dollars, amounting to $30 billion.

ADS-equipped vehicles could also expand transportation access to underserved populations, such as people with disabilities and communities lacking access to transportation.

By communicating with one another and with infrastructure, autonomous vehicles could also operate more efficiently, reducing environmental costs and saving passengers hours spent in traffic; a 2014 analysis estimated that Americans spent upwards of 6.9 billion hours stuck in traffic that year.

ADSs add weight to cars, which can increase greenhouse gas emissions. However, a 2018 study found that the efficiency improvements provided by small- and medium-sized ADSs could lead to a 6-9% overall reduction in greenhouse gas emissions over a vehicle's lifetime.

ADASs cover a broad range of technologies that are designed to aid drivers, many of which are included in modern vehicles manufactured today. The four specific systems that the NHTSA recommends are: forward collision warning, lane departure warning, crash imminent braking, and dynamic brake support.

Researchers have attempted to determine the societal benefits of ADASs. One 2023 study estimated that between 2021 and 2050, the use of ADAS technologies in the United States could prevent around 37 million crashes, averting 14 million injuries and almost 250,000 fatalities.

Autonomous vehicle vs. non-automated vehicle crashes

In 2021, there were 42,939 people killed in traffic crashes in the U.S., 10% more than in 2020. Many of these crashes were linked to human error. Fatalities due to alcohol impairment rose 14% between 2020 and 2021. Roadway injuries increased from 2.28 million in 2020 to 2.50 million in 2021. The total number of traffic crashes reported by police in 2020 was 5.25 million; this increased to 6.1 million in 2021. In 2021, Americans drove a total of 3.14 trillion miles.

U.S. traffic crash statistics (reported in thousands)

Source: National Highway Traffic Safety Administration (NHTSA); *Indicates statistically significant change with a significance level of α=.05

In June 2021, the NHTSA issued a Standing General Order requiring timely crash reports from any manufacturers or operators using automated systems categorized as SAE Level 2 or higher. These reporting requirements are more stringent than typical crash reporting. A report must be filed if a vehicle’s ADAS or ADS was engaged within 30 seconds of the incident.

Between January 1 and December 15, 2023, the NHTSA documented 269 ADS crashes. Since reporting began in July 2021, there were 508 ADS crashes reported.

Two of the largest operators of ADS-equipped vehicles in the U.S. are Waymo LLC – owned by Google’s parent company, Alphabet – and Cruise LLC – owned by General Motors. Since July 2021, Waymo has reported 215 crashes, and Cruise has reported 151. California, the hub of ADS road testing, was home to 325 of these crashes.

No injuries were reported in 439 of these crashes, four resulted in a serious injury, and none resulted in a fatality.

However, this does not imply that ADS systems are not lethal. In July 2023, a guilty verdict was reached for the 2018 death of a pedestrian who was struck by a fully autonomous Uber in Phoenix, AZ. The “backup driver”, whose job was to monitor the road and take control when needed, was accused of being distracted by her phone when the pedestrian crossed her path.

Comparing human and ADS safety

When considering the safety of autonomous vehicles, it’s important to ask what they are safe in comparison to.

Comparing humans and autonomous vehicles is methodologically and statistically challenging. One issue is that there are far fewer autonomous vehicles than human drivers. Another issue is that the crash reporting requirements are far stricter for ADS-equipped vehicles.

The NHTSA's estimates for human driver crash rates are calculated using a sampling of police crash reports, which do not include unreported crashes, and fatal crash statistics. Due to stricter reporting, the NHTSA’s data on automation-related crashes are likely more accurate, leading some researchers to use alternative data sources for calculating human crash rate estimates.

Fully autonomous vehicles are often barred from driving on certain roads, such as highways, meaning that their crash statistics are geographically different than human drivers.

One 2016 study, commissioned by Google and carried out by researchers at the Virginia Tech Transportation Institute, compared human-driver crash rates to those of Google’s self-driving cars. At the time, Google’s self-driving car crash data only included 16 crashes.

Significant differences between human and self-driving car crash rates were only found for crashes resulting in minor damage. For these crashes, the human rate was 14.4 crashes per million miles driven, while self-driving cars had a crash rate of 5.6. However, the sample sizes between the two groups were extremely skewed.

A 2023 study – carried out by researchers from Swiss Reinsurance Company and Waymo, an autonomous vehicle ride-hailing company – compared accident insurance claims between humans and autonomous vehicles.

The researchers found that Waymo vehicles, operating fully autonomously, produced zero bodily injury claims per million miles driven compared to the 1.11 claim rate for human drivers. The study also notes that autonomous vehicles drove for 3.8 million miles with no human behind the wheel with zero bodily injury claims.

Property damage claims were also significantly lower for Waymo autonomous vehicles, which incurred 0.78 of these claims per million miles compared to 3.26 for the human drivers.

Researchers are also trying to understand whether humans and self-driving cars crash for different reasons. A 2021 study comparing crash rates between human drivers and self-driving cars limited their analysis to crashes where the vehicle was struck from behind.

The researchers found that self-driving cars were 4.8 times more likely to be struck from behind than human drivers.

The self-driving cars were also significantly more likely than human drivers to be struck while stopped. The researchers hypothesized that self-driving vehiclesmay be stopping unexpectedly.

ADAS crash statistics

Given that the NHTSA’s Standing General Order includes reporting requirements for vehicles using SAE Level 2 systems, crash statistics are also available for these ADASs, such as Tesla’s Autopilot system.

Between January 1 and December 15, 2023, there were 475 crashes reported involving a vehicle using a Level 2 ADAS system. Since reporting began in July 2021, there have been 1,239 Level 2 ADAS-involved crashes documented.9 Tesla accounted for most of these, racking up 1,047 crashes.

Level 2 ADASs were involved in at least 15 crashes resulting in serious injury and 28 fatal crashes. However, 1,061 crashes were reported with an unknown injury level. This means that in a majority of Tesla's crash reports, this information is left blank. A Washington Post analysis showed that Tesla's Level 2 ADAS was involved in 17 fatal crashes between 2019 and May of 2023.

During that same time period, 3.5% of crashes resulted in injuries that were deemed moderate or severe, while 83.2% were listed as unknown.

Level 2 ADASs require the driver to remain vigilant for any potential hazards and stay prepared to take over control if the system can't operate safely. However, these systems also allow drivers to multitask while the ADAS is engaged, leading researchers to study how multitasking can affect a driver's readiness for dangerous situations.

One 2017 study put participants in a Level 2 ADAS driving simulator, which would suddenly ask them to take over driving. Half of the participants were asked to multitask by reading a newspaper while waiting to resume control; the other half remained fully focused on the simulator. The researchers found that participants who were reading took around 6.06 seconds to resume control while the non-reading group took only around 4.46 seconds.

A similar study involving 72 participants found similar results in a secondary analysis when the multitasking group was asked to simulate a phone call.

The future of autonomous vehicles

Adopting these automated systems will require public trust in their safety. Studies examining trust in automated systems have found conflicting results, some showing higher levels of trust and others showing greater distrust.

A 2021 report from the American Automobile Association Foundation for Traffic Safety examined survey respondents’ opinions regarding automated vehicles. The researchers found that trust in these vehicles generally decreased as the level of automation increased.

When asked about crash prevention, respondents expressed more trust in Level 2 and 3 systems than Levels 4 and 5. This finding also varied across age groups. For example, respondents aged 25 to 39 trusted Level 5 systems more than senior citizens did.

Levels of trust also differed across genders. For example, 56% of both men and women said they either strongly or somewhat trusted Level 2 systems. When asked about Level 5 systems, 42% of men and only 29% of women showed that amount of trust.

When asked which level of automation they would like to own, if money was of no concern, the highest number of respondents said they would be most comfortable owning a vehicle with SAE Levels 0 or 1 (at 28%). Respondents showed the least comfort in owning a vehicle equipped with Level 4 or 5 systems (12% and 17%).


How many self-driving car crashes happen in the United States?

Since official reporting began in July 2021, the National Highway Traffic Safety Administration (NHTSA) has reported 508 crashes involving vehicles driving autonomously. Consumer automobiles operating autonomously, such as Teslas in Autopilot mode, have been in 1,239 crashes in that time.

Can consumers purchase self-driving cars?

Consumers cannot currently purchase fully autonomous vehicles equipped with an Automated Driving System (ADS). Consumers can purchase vehicles equipped with an Advanced Driver-Assistance System (ADAS), some of which can control steering, acceleration, and braking simultaneously.

Are self-driving cars safer than human drivers?

It’s still too early to definitively say whether self-driving cars are safer than human drivers. Waymo and Cruise are already operating fully autonomous ride-hailing services in certain cities, and several studies have reported that these autonomous vehicles have a lower crash rate than human drivers.


  1. “Autonomous Vehicles Factsheet.” University of Michigan Center for Sustainable Systems. Evaluated Jan. 30, 2024. Link Here
  2. Shi, E., et al. “The Principles of Operation Framework: A Comprehensive Classification Concept for Automated Driving Functions.” Society of Automotive Engineers (SAE) International Journal of Connected and Automated Vehicles. Evaluated Jan. 30, 2024. Link Here
  3. “Preparing for the Future of Transportation.” U.S. Department of Transportation. Evaluated Jan. 30, 2024. Link Here
  4. “Automated Vehicles for Safety.” National Highway Traffic Safety Administration (NHTSA). Evaluated Jan. 30, 2024. Link Here
  5. Gawron, J. H. et al. “Life Cycle Assessment of Connected and Automated Vehicles: Sensing and Computing Subsystem and Vehicle Level Effects.” Environmental Science & Technology. Evaluated Jan. 30, 2024. Link Here
  6. “Driver Assistance Technologies.” National Highway Traffic Safety Administration (NHTSA). Evaluated Jan. 30, 2024. Link Here
  7. “Examining the Safety Benefits of Partial Vehicle Automation Technologies in an Uncertain Future.” American Automobile Association (AAA) Foundation for Traffic Safety. Evaluated Jan. 30, 2024. Link Here
  8. “Overview of Motor Vehicle Traffic Crashes in 2021.” National Highway Traffic Safety Administration (NHTSA). Evaluated Jan. 30, 2024. Link Here
  9. “Standing General Order on Crash Reporting.” National Highway Traffic Safety Administration (NHTSA). Evaluated Jan. 30, 2024. Link Here
  10. Billeaud, J. & Snow, A. “The backup driver in the 1st death by a fully autonomous car pleads guilty to endangerment.” Associated Press. Evaluated Jan. 30, 2024. Link Here
  11. Blanco, M. et al. “Automated Vehicle Crash Rate Comparison Using Naturalistic Data.” Virginia Tech Transportation Institute. Evaluated Jan. 30, 2024. Link Here
  12. Di Lillo, L. et al. “Comparative Safety Performance of Autonomous- and Human Drivers: A Real-World Case Study of the Waymo One Service.” Preprint at Evaluated Jan. 30, 2024. Link Here
  13. Goodall, N. J. “Comparison of automated vehicle struck-from-behind crash rates with national rates using naturalistic data.” Accident Analysis & Prevention. Evaluated Jan. 30, 2024. Link Here
  14. Siddiqui, F. & Merrill, J. B. “17 fatalities, 736 crashes: The shocking toll of Tesla’s Autopilot.” Washington Post. Evaluated Jan. 30, 2024. Link Here
  15. Eriksson, A. & Stanton, N. A. “Takeover Time in Highly Automated Vehicles: Noncritical Transitions to and From Manual Control.” Human Factors: The Journal of the Human Factors and Ergonomics Society. Evaluated Jan. 30, 2024. Link Here
  16. Gold, C. et al. “Taking Over Control From Highly Automated Vehicles in Complex Traffic Situations: The Role of Traffic Density.” Human Factors: The Journal of the Human Factors and Ergonomics Society. Evaluated Jan. 30, 2024. Link Here
  17. “Users’ Trust in and Concerns about Automated Driving Systems.” American Automobile Association (AAA) Foundation for Traffic Safety. Evaluated Jan. 30, 2024. Link Here
  18. “The Economic and Societal Impact of Motor Vehicle Crashes, 2019 (Revised).” National Highway Traffic Safety Administration (NHTSA). Evaluated Feb. 12, 2024. Link Here


Back to ConsumerAffairs

Journal of Consumer Research