Self-Driving Car Crashes Triple in 2025: What to Know Before Buying One

Driverless Cars Aren’t Just Crashing in Bad Weather. Here’s What Buyers Should Know in 2025

Advanced Driver Assistance Systems (ADAS) and full-fledged Automated Driving Systems (ADS) now show up everywhere from family SUVs in the countryside to driverless robo-taxis weaving through busy city streets.

The widespread adoption of such technology has surged faster than safety mechanisms can be put in place. This raises a key concern: How often do these systems crash and what can be done to ensure safety guardrails keep pace with the explosive growth of tech advancement?

Using newly released crash data from the National Highway Traffic Safety Administration (NHTSA), Bumper analyzed 2,150 reported incidents involving vehicles equipped with ADAS or ADS technology. The study explores when and where crashes occur, which manufacturers appear most frequently in reports, and what all of this means for shoppers considering a used vehicle with semi-autonomous features.

Key findings at a glance

Most incidents were minor, not catastrophic

About 82% of the incidents were simple property-damage events—the automotive equivalent of stubbing your toe. These are common incidents you’d usually see in dense urban traffic or crowded intersections.

• 81.4% ADAS (1752 vehicles)
• 18.5% ADS (398 vehicles)

The data suggests that while these systems aren’t flawless, most issues don’t escalate into serious injury situations.

California is the epicenter

More than half of all reported cases occurred in California, and San Francisco alone accounts for 42% of incidents reported nationally. The city’s combination of steep streets, dense traffic, and heavy ADS testing makes it a natural hotspot for incidents.

Crashes surged beginning in 2024 and climbed sharply in 2025

2025 opened with more than triple the number of reports logged in any earlier year.

This jump lines up with wider ADS deployment, expanded data-reporting requirements, and more used vehicles entering the market with partially automated features that everyday drivers may not fully understand yet.

Waymo and Tesla dominate the numbers

Together, Waymo and Tesla account for 58.74% of all reports. That doesn’t automatically brand them as unsafe—it more accurately reflects their enormous presence in both autonomous fleets (Waymo) and consumer-level driver-assist systems (Tesla). More vehicles and more miles inevitably mean more reports.

Where, when, and how crashes occur

Out of the 2,150 total reported incidents, most were relatively minor, but the data still shows some patterns worth paying attention to, especially if you’re evaluating a used vehicle with advanced automation.

What the crashes look like

• Property damage only — 1,764 (82.04%)

These are the everyday bumps that happen in cramped traffic or while maneuvering at low speeds. A share came from heavy-duty vehicles—more than 15% of all property-damage events involved large vans or trucks equipped with ADAS, which tend to operate in delivery fleets and rack up significant mileage.

• Injury-related crashes 330 — (15.3%)

Injury cases remain a minority, though the numbers rise slightly in cities where ADS vehicles operate around the clock.

• Fatalities — 56 (2.6%)

Fatal incidents are rare but notable. They most commonly involved:

• Tesla: 46
• Ford: 3

In most cases, these reflect high-speed conditions or situations where automation was not the only factor at play. NHTSA’s reporting rules also mean that manufacturers often file a report even when automation wasn’t active at the moment of impact.

Incident trends over time

Between 2020 and 2023, reports stayed relatively low, typically 20 to 40 per year. Automated systems were still in limited circulation, and only a few states required detailed reporting.

That picture changed quickly in 2024, when incident totals began climbing into monthly highs near 80. By early 2025, the system saw an unprecedented surge that dwarfed the previous years combined.

Several factors contributed to the jump:

• Rapid expansion of autonomous taxi and delivery fleets
• More mainstream cars offering ADAS as standard equipment
• Stricter state reporting rules, particularly in high-usage regions like California
• Higher overall mileage as automated vehicles began operating longer hours and in denser areas

When crashes happen

Time-of-day patterns are surprisingly consistent:

• Safest period: Early morning, when traffic is light and sensors face fewer obstructions.
• Highest-risk period: Late evening, when visibility drops and traffic grows unpredictable.

Where crashes are happening

Top states for ADAS/ADS incidents

1. California – 1,231 (57%)
2. Arizona – 342 (15.9%)
3. Texas – 221 (10.27%)
4. Nevada – 49 (2.2%)
5. Florida – 47 (2.1%)

These states account for more than 85% of all reported incidents, largely because they host the bulk of driverless testing programs and early consumer adoption.

Top cities

• San Francisco – 894
• Phoenix – 156
• Los Angeles – 143
• Austin – 119
• Tempe – 96

San Francisco stands out dramatically, thanks to its dense urban layout and the high concentration of ADS fleets operating daily. Other cities on the list are major hubs for either autonomous taxis or large-scale ADAS-equipped delivery operations.

ADS vs. ADAS by city

A look at city-level data shows just how unevenly automated driving systems are distributed across the country. Some metro areas are almost entirely dominated by driverless (ADS) fleets, while others see a heavier mix of traditional consumer vehicles using ADAS features.

These ratios make one thing clear: cities like San Francisco, Phoenix, and Tempe are ground zero for driverless fleets. Waymo and Cruise operate thousands of autonomous miles in these locations every day, and their 24/7 fleet activity naturally leads to dramatically higher ADS reporting.

For shoppers, this context matters: ADAS-equipped consumer vehicles simply don’t rack up the same nonstop mileage or ultra-dense urban exposure that ADS fleets do.

Manufacturer insights

Who reports the most incidents?

A handful of manufacturers dominate the data, though largely because they operate the largest fleets or maintain the strictest reporting pipelines. Out of the 2,150 reports Bumper analyzed, these brands were reported the most:

• Waymo — 1,080 reports (50.23%)
• Tesla — 183 (8.51%)
• General Motors — 164 (7.63%)
• Cruise LLC — 149 (6.93%)
• Transdev Alternative Services - 133 (6.18%)
• Zoox - 75 (3.4%)
• Sabaru of America - 49 (2.27%)
• Ford Motor Company - 45 (2.09%)
• Honda - 33 (1.53%)
• Argo AI - 20 (0.93%)

Taken together, Waymo and Tesla represent 58.5% of all crash reports.

While these figures may look dramatic, they’re often more reflective of scale than safety. Waymo’s fleet logs enormous mileage in dense cities, and Tesla has millions of ADAS-equipped consumer vehicles on the road.

Crash types by manufacturer

• Waymo: Nearly all incidents—1,012 of 1,080—were property-damage crashes. Most involve low-speed scrapes with parked cars or delivery vehicles in dense environments.
• Tesla: Recorded the highest number of fatality-related reports (46), though many involve scenarios where Autopilot or FSD may not have been the primary factor.
• Ford: Low overall volume, but includes three fatality reports, putting it among the top non-Tesla contributors to fatal incident filings.

Road conditions when crashes occur

There’s a common assumption that automated systems turn risky when the weather is unfavorable. Strangely enough, the data points in the opposite direction. Most incidents happen when the skies are clear, the pavement is dry, and visibility is good—conditions that should be the easiest for sensors and software.

Lighting

• Daylight: 1,159
• Dark: 642

Even though ADS vehicles rely on lidar and radar, daylight still appears to be their busiest—and riskiest—operating window. Evening hours remain challenging as traffic patterns shift and visibility becomes more complex for both sensors and human drivers sharing the road.

Road type

• Streets: 991
• Intersections: 634
• Highways: 279
• Parking lots: 121

Urban streets and intersections account for the bulk of incidents. These environments produce a constant flow of unpredictable variables—pedestrians stepping into crosswalks, tight delivery zones, cyclists weaving through traffic, and dense clusters of parked cars. Highways, though high-speed, tend to be more structured and thus easier for automation.

Weather

• Clear: 1,542
• Cloudy: 284
• Rain: 68

Clear conditions dominate, partly because most ADS fleets avoid operating in heavy rain by design. When the weather worsens, fleets either reduce service or rely more heavily on remote operators, which limits exposure to higher-risk situations.

What it means

Taken together, the data suggests that ADAS and ADS systems currently struggle more with traffic complexity than with weather itself. The hardest challenges aren’t storms or snow—they’re crowded streets, busy intersections, and the countless split-second decisions happening around the vehicle.

What this means for used-car buyers

Advanced driver-assist features can be exciting upgrades, but—like any technology—they only work as well as the hardware, software, and maintenance behind them. If you’re considering a used car with ADAS features such as adaptive cruise control, lane keeping, or automatic emergency braking, here are the key things to keep in mind.

Not all ADAS systems are equal

Some manufacturers (Subaru, Honda) show higher injury-related crash rates in the dataset. That doesn’t necessarily make them unsafe, but used buyers should:

• Test sensors and cameras during the drive
• Confirm the system has been properly calibrated after past repairs
• Check for open recalls
• Review crash and repair history with a VIN lookup

*Vehicle reports are based on data available and may not include historical accident records in all states.

A well-maintained ADAS system can be an asset—one that quietly works in the background to reduce everyday driving risks.

ADS (driverless) activity matters even if you’re not buying an autonomous car

Cities with high ADS activity (San Francisco, Phoenix, Austin) may see more interactions between human-driven and autonomous cars. These areas also see more low-speed, property-damage incidents involving ADS fleets.

If you’re buying in these areas:

• Inspect for subtle bumper or corner-panel impacts
• Look for notes related to ADAS or ADS involvement in Bumper’s vehicle history tools
• Ask sellers about prior repairs to sensors, cameras, or radars

Minor scrapes from automated fleet interactions may not show up on standard accident reports, but they can still affect how well sensors work.

ADAS sensors can be expensive to repair

One of the biggest surprises for many buyers is how costly even a small fender-bender can be. Modern driver-assist systems rely on carefully aligned sensors that don’t work well with even minorly damaged machinery.

Potential expenses include:

• Radar-unit replacements
• Camera recalibration
• Bumper or grille sensor realignment

Before buying, verify:

• All sensors are functional during the test drive
• Calibration records exist for any previous repairs
• Aftermarket body work hasn’t interfered with sensor placement

A used car with misaligned sensors may behave unpredictably, sometimes without obvious warning messages.

Check the car’s crash history carefully

Because 82% of reported incidents are minor property-damage events, not all of them end up in traditional accident databases. Still, these small hits can undermine sensor reliability, so it is essential to run a vehicle history report.

When reviewing a report, pay extra attention to repairs involving:

• Front or rear bumpers
• Quarter panels
• Headlights or taillights (common sensor locations)
• Windshield replacements (for camera-based systems like Subaru EyeSight or Honda Sensing)

Any work in these areas should come with documentation showing proper calibration afterward.

Don’t overestimate automation

The crash data makes one thing clear: ADAS is a support system, not a self-driving solution. Most incidents occur in busy city environments and during evening hours—times when driver attention matters most.

If the used car is marketed as “semi-autonomous” or comes with advanced features like highway hands-off modes:

• Review the owner’s manual
• Ask the seller to demonstrate features
• Test them safely on your drive
• Understand exactly when the system disengages or requires hands-on control

These features can help, but they’re only as effective as the driver using them.

Final takeaways

Self-driving and driver-assist systems are expanding rapidly, and the surge in incident reports—especially in 2025—reflects that growth. Most incidents involve only minor damage, but they also highlight the real-world limits of automation in dense urban traffic.

For used-car shoppers, ADAS-equipped vehicles can offer genuine benefits. They can reduce crash severity, shorten braking distances, and provide an extra layer of awareness. But they also require careful inspection, proper calibration, and realistic expectations, especially when you’re buying used.

Bottom line: ADAS and ADS vehicles aren’t inherently dangerous, but they’re not flawless. Understanding how these systems perform—and what they might cost to maintain—can help you make a smarter, safer used-car purchase.