Digital distractions & road safety
How often do self-driving cars crash?
5,202 crashes involving self-driving or driver-assist cars have been reported to regulators since mid-2021, and Tesla accounts for more than half of them (2,732), because it has the largest fleet using Level 2 ADAS. Reporting became mandatory in June 2021, and the yearly count has grown from 329 incidents that year to 1,384 in 2024. Waymo leads among true autonomous systems with 1,443 reported incidents but also logs the most autonomous miles. Of all incidents, 7.4% caused injury and 1.2% were fatal. The rising count reflects expanding fleets and better compliance rather than a deteriorating safety record.
Roughly 5.6 crashes per day.
Source: NHTSA Standing General Order (SGO) AV/ADAS database, through November 2025.
Last checked: 18 October 2026
What self-driving car crashes mean for you as a future passenger and pedestrian
If you have ridden in a Tesla on Autopilot, a Waymo robotaxi, or any vehicle with advanced driver assistance, you have direct experience with this technology. NHTSA's Standing General Order database captures every incident where automation was active, from a Tesla Autopilot fender-bender in a parking lot to a Waymo collision. It is the only systematic public record of what goes wrong.
The safety comparison to human drivers is genuinely complicated. Waymo has published internal data suggesting a significantly lower serious injury rate per mile than the US average. Tesla's safety reports make a similar claim for Autopilot-active miles. But both comparisons have a methodological problem: AVs primarily operate in clear conditions on well-mapped roads, sidestepping the scenarios that account for most human crashes.
For pedestrians and cyclists, the picture is less settled. Autonomous vehicles have killed pedestrians, most clearly in the Uber test-vehicle fatality of 2018. Early robotaxi incidents in San Francisco raised questions about edge-case handling. The technology is improving, and will likely deploy at meaningful scale within this decade. Whether that is reassuring depends entirely on which baseline you use.
AV crash statistics: what the NHTSA data shows
NHTSA SGO cumulative total: 5,202 AV/ADAS incidents reported through November 17, 2025
2024 saw 1,384 AV incidents, nearly double the 742 recorded in 2022
7.4% of AV incidents resulted in injury; 1.2% resulted in fatality
Tesla accounts for 2,732 incidents (mostly Level 2 ADAS); Waymo accounts for 1,443 (Level 4 ADS)
In 2024, NHTSA opened a preliminary investigation into Waymo after 22 incidents involving single-party crashes and potential traffic control device violations
AV incidents vs. human distraction deaths, today
The scales here are very different. Human distraction kills far more people every day than AV incidents cause, which is the baseline the industry points to when making its safety case.
From Google's first test car to fatal crashes: AV history
- 2018Uber AV fatally strikes pedestrian in Tempe, Arizona; first AV pedestrian death in history
- 2021NHTSA issues Standing General Order; mandatory crash reporting for ADS/ADAS begins; 329 incidents reported
- 2022NHTSA SGO: 742 incidents; first full year of systematic AV crash data
- 2023Cruise ADS drags pedestrian about 20 feet; Cruise suspends driverless operations; 921 NHTSA incidents in 2023
- 20241,384 incidents; NHTSA opens Waymo preliminary investigation; Waymo expands to multiple cities
- 20251,793 incidents through Nov 17; cumulative total reaches 5,202
AV/ADAS incident reports over time
NHTSA mandatory AV crash reporting began in 2021 and has documented steady growth from 329 incidents in 2021 to 1,384 in 2024, driven by expanding fleets, improving compliance, and the entry of more operators into the market.
| Year | Rate | Incidents/year | Context |
|---|---|---|---|
| 2021 | 2.00/day | 329 | First year of mandatory NHTSA reporting |
| 2022 | 2.00/day | 742 | Rapid increase as fleet grows and compliance improves |
| 2023 | 2.50/day | 921 | Growth continues; regulatory scrutiny increases |
| 2024 | 3.80/day | 1.4K | Continued expansion of AV fleets; Waymo major expansion |
| 2025 (partial) | 5.60/day | 1.8K | On track for ~2,000+ for full year |
| 2027 (forecast) | 9.00/day | 3.3K | AV fleet expansion; Waymo, Tesla, new entrants add miles |
Self-driving crash data: what NHTSA's database actually tells us
What the numbers actually mean
Each NHTSA incident in the database means only that automation was active within 30 seconds of a crash, not that automation caused it. That includes trivial fender-benders where a human rear-ended an AV stopped at a red light. NHTSA collects the data to detect patterns, not to assign blame. Waymo has been transparent that the majority of its reported incidents involve being struck by human-driven vehicles, not failures of its own system.
The future of AV safety data
Per-mile-driven metrics, not raw incident counts, are where this comparison is headed. Early data from Waymo's public reports puts its at-fault injury collision rate below the human baseline per million miles. The caveat: AVs currently operate mostly in clear conditions on well-mapped streets, sidestepping the scenarios that cause most human crashes. Edge cases are what the regulator cares about. The 2023 Cruise incident, where a pedestrian was dragged 20 feet by a vehicle that failed to stop, is the clearest example of what a failure mode looks like.
Research & incident data from NHTSA and industry
| Year | Finding | Value | Source |
|---|---|---|---|
| 2022 | NHTSA SGO Report June 2022: First formal AV crash reporting summary; 329 incidents in 2021 under new mandatory reporting | 329 incidents 2021 | NHTSA |
| 2022 | NHTSA SGO 2022 full year: 742 AV incidents total; NHTSA's first release (July 2021 to May 2022) counted 392 Level 2 ADAS crashes (Tesla 273) and 130 ADS crashes (Waymo 62) | 742 incidents 2022 | NHTSA |
| 2023 | NHTSA SGO 2023 full year: 921 incidents; Cruise suspends robotaxi operations after pedestrian dragging incident | 921 incidents 2023 | NHTSA |
| 2024 | NHTSA SGO 2024 full year: 1,384 incidents; Waymo expands to multiple US cities; PE opened for Waymo ADS | 1.4K incidents 2024 | NHTSA |
| 2025 | NHTSA SGO through Nov 17, 2025: 1,793 incidents in under 12 months; on track for ~2,000+ full year | 1.8K incidents Jan-Nov 2025 | NHTSA |
In perspective
At the 2024 rate of 3.8 incidents/day, there is a reported AV system incident somewhere in the US roughly every six hours.
5,202 total incidents through late 2025; 1.2% involved a fatality. That is about 62 cases. Not all were caused by the automation system: many involved the AV being struck by a human driver.
Frequently asked questions
- How many self-driving car crashes have been reported to NHTSA?
- As of November 17, 2025, NHTSA had received 5,202 total incident reports under the Standing General Order, which covers SAE Level 2+ automated systems. The count has grown from 329 in 2021 to 1,793 in the first 10.5 months of 2025 alone.
- Does the rising crash count mean self-driving cars are getting more dangerous?
- Not necessarily. The increase is driven by three factors: (1) growing fleet size and miles driven, (2) greater compliance with reporting requirements over time, and (3) more companies entering the market. On a per-mile-driven basis, advanced AV systems like Waymo have claimed safety records comparable to or better than human drivers, though independent verification remains challenging.
- Which self-driving car company has had the most crashes?
- Tesla has the most total incidents (2,732) because it has by far the largest fleet using ADAS (Level 2) systems like Autopilot. Among true ADS (Level 3-5) systems, Waymo leads with 1,443 reported incidents, but Waymo also drives the most ADS miles in the US. Cruise (GM) had 155 ADS incidents; it paused driverless operations after a 2023 pedestrian incident, and GM ended the robotaxi program in December 2024.
Cite this rate
AnythingCounter derived rate, last checked October 2026: about 5.6 crashes per day.
Method: Based on NHTSA 2025, shown as a daily rate.
Suggested citation: AnythingCounter, "Crashes caused by self-driving cars," October 2026, https://anythingcounter.com/self-driving-crash-statistics
Primary source: NHTSA.
How this number is calculated, and how much to trust it
1,384 US AV/ADAS crashes in 2024 ÷ 365 days ≈ 3.8 incidents/day ÷ 86,400 seconds ≈ 0.000044/sec. The live counter shows cumulative 2024-rate incidents today. Importantly, this figure is not comparable to conventional crash statistics: it includes Level 2 ADAS events (Tesla Autopilot) that involve large fleet sizes and the reporting is mandatory for automation-active incidents, not a measure of AV safety vs. conventional driving.
All data comes from NHTSA's publicly downloadable SGO database, filed directly by manufacturers and operators. It is the only mandatory, systematic record of automated-driving-system incidents anywhere. Reports are filed within 10 days of an incident; NHTSA updates the dataset monthly. The program began June 29, 2021.
Sources: see below. More detail on our methodology page.
Worth a look... Distracted driving deaths - Distraction accidents - Robots deployed.