Los Angeles is becoming one of America’s biggest real-world tests of driverless transportation, and the growing fleet is producing a growing crash record. Waymo has reported 186 crashes in Los Angeles County so far in 2026, after 219 in 2025 and 69 in 2024, according to government crash data analyzed by UC Berkeley researchers and the Los Angeles Times.
The raw total sounds alarming until the incidents are examined more closely. More than 80% of this year’s reported crashes produced no apparent injuries, and dozens happened while the Waymo vehicle was stopped. At the same time, several incidents involving pedestrians and injured motorists show why regulators are paying close attention. The emerging picture is more complicated than either “robotaxis are dangerous” or “robots are safer”: Waymo is driving vastly more miles, reporting many minor collisions, and accumulating the evidence needed to answer the safety question more convincingly.
The Crash Count Has Risen Quickly
Waymo reported 69 Los Angeles County crashes in 2024, followed by 219 in 2025 and another 186 so far in 2026. On the surface, that trajectory looks dramatic. The 2025 total was more than triple the previous year’s figure, and the current year had already accumulated hundreds of incidents before reaching its final months. The numbers reflect collisions reported under federal autonomous-vehicle requirements and compiled by UC Berkeley’s Safe Transportation Research and Education Center.
The missing context is exposure. Waymo is no longer operating a tiny experimental fleet making occasional trips through a narrowly defined part of Los Angeles. The company has expanded its commercial service and accumulated tens of millions of driverless miles in the region. That means substantially more opportunities to encounter distracted motorists, parking-lot bumps, rear-end collisions, debris, pedestrians and other routine hazards. An expanding crash count can therefore coexist with an improving crash rate. For safety purposes, the number of incidents matters, but the number of miles driven matters just as much.
More Than 80% Produced No Apparent Injuries
The injury breakdown changes the tone of the 186-crash figure considerably. Just over 80% of the Waymo crashes reported in Los Angeles County this year resulted in no apparent injury, according to the latest review of the data. Approximately 15% were classified as involving a possible injury, while fewer than 1% were categorized as involving a suspected serious injury. The dataset therefore contains a large number of incidents that would typically be described as minor collisions rather than major wrecks.
That does not make those crashes meaningless. Property damage still costs money, passengers can find even a low-speed collision frightening, and recurring minor incidents may reveal situations an automated system needs to handle better. Yet crash severity matters when judging whether a transportation technology is making roads more or less dangerous. A system involved in many low-speed contacts but comparatively few serious crashes presents a different public-safety profile from one producing fewer reported collisions but more hospitalizations. The headline count is useful; the injury distribution provides the context needed to understand it.
Many Waymos Were Not Even Moving When They Were Hit
One of the more surprising details in the Los Angeles data is how frequently Waymo vehicles were stationary. Of the 186 reported crashes this year, 82 occurred while the autonomous vehicle was stopped. Another 65 happened while the Waymo was traveling straight. The crash reports include incidents in which other motorists struck stationary robotaxis, emphasizing why an involvement count cannot automatically be interpreted as a fault count.
One federal report described a Waymo that had stopped for a passenger drop-off on a narrow Los Angeles street with its hazard lights activated. Another vehicle turned into a driveway, reversed back toward the roadway and contacted the stationary Waymo. In a separate example reviewed by the Los Angeles Times, a Waymo waiting at a traffic light was struck from behind by a truck. Such cases resemble the frustrations experienced by ordinary motorists who are rear-ended at a red light. For an autonomous fleet, however, they become part of a highly visible government database and consequently part of the broader debate over robotaxi safety.
Most Collisions Still Involve Ordinary Passenger Vehicles
Passenger cars are by far the most common other party in this year’s Los Angeles Waymo crashes. Of the incidents reviewed, 149 involved a passenger vehicle. Eleven involved a heavy truck, while two involved first-responder vehicles. The mix is important because robotaxis operate within a transportation system still overwhelmingly controlled by humans. Even a perfectly functioning automated car would continue to encounter mistakes made by surrounding motorists.
One reported collision illustrates the complexity. A Waymo moved aside to accommodate an emergency vehicle but was subsequently struck by that vehicle. No injuries were reported, although the autonomous car required towing. Cases such as this complicate attempts to assign a simple meaning to a robotaxi crash total. An autonomous vehicle can cause a collision, contribute to one, fail to avoid someone else’s mistake or merely be the object another vehicle hits. Determining whether robotaxis are genuinely safer therefore requires examining responsibility, severity and mileage rather than treating every reported contact as an identical failure of the automated-driving system.
Federal Rules Capture Crashes Human Drivers May Never Report
Waymo’s crash record is unusually visible because federal rules require companies operating automated driving systems to report qualifying incidents to the National Highway Traffic Safety Administration. Under the current requirements, an incident can be reportable when the automated system was operating within 30 seconds of a crash and the collision met specified property-damage or injury criteria. Vulnerable-road-user impacts, airbag deployments, tow-away crashes, fatalities and certain medically treated injuries are among the events covered.
That creates an important comparison problem. Autonomous vehicles generally have extensive cameras, sensors, telemetry and company oversight, meaning an operator may know almost immediately when a collision occurs. NHTSA explicitly warns that manufacturers with better data collection can appear to have more incidents simply because they know about more of them. Many tiny human-driver collisions are never reported to police or federal authorities. A low-speed scrape that becomes a data point in Waymo’s federal file might have ended between two human drivers with an exchange of insurance information—or no formal record whatsoever.
More Cars and Miles Make Raw Crash Totals Less Useful
Waymo’s Los Angeles exposure has grown enormously. The company reports approximately 67.1 million rider-only miles in the Los Angeles market through June 2026. Nationwide, Waymo said earlier this year that it was providing roughly 500,000 paid rides per week across its growing network of cities. California regulators have also approved a much broader expansion footprint, potentially bringing autonomous service into additional communities as deployment proceeds gradually.
The Los Angeles Times found that Waymo’s reported crash rate has moved in the opposite direction from its raw collision count. Its analysis put the rate at roughly 57 reported crashes per million miles in 2023, compared with fewer than six per million miles in 2026. That is a crucial distinction. If a fleet drives ten times farther, it can accumulate more crashes while becoming safer for every mile travelled. NHTSA itself cautions against comparing raw crash totals without exposure data because operators run different numbers of vehicles, in different places and for dramatically different distances. Scale changes the denominator.
Waymo’s Safety Data Show Much Lower Injury Rates
Waymo argues that its driverless system performs considerably better than human drivers when comparable mileage and operating environments are considered. Through June, its safety dashboard reported an any-injury crash rate in Los Angeles of 0.86 incidents per million rider-only miles, compared with a human benchmark of 2.55. For crashes producing a serious injury or worse, Waymo reported zero incidents per million miles in Los Angeles over that dataset, compared with a modeled human benchmark of 0.16.
Those figures deserve both attention and qualification. They come from Waymo’s own safety analysis, although the underlying crash reports are tied to federal reporting data and the company publishes its methodology and downloadable datasets. The human benchmark is adjusted to better reflect the locations and road types where Waymo actually operates rather than comparing robotaxis with every kind of American driving. That improves the comparison, but it also underlines an important limitation: a Waymo driving urban streets within a defined operating area is not being tested against every condition a human might encounter, such as every freeway, snowstorm or remote rural road.
Peer-Reviewed Research Also Points Toward a Safety Advantage
The safety argument is not based solely on Waymo’s current website. A 2025 peer-reviewed study in the journal Traffic Injury Prevention examined 56.7 million rider-only Waymo miles and compared crashes with human-driving benchmarks aligned by location, road and vehicle type. Across the full dataset, the researchers found statistically significant reductions in crashes involving reported injuries, airbag deployments and suspected serious injuries. Intersection crashes produced particularly large reductions.
There is an important disclosure: the researchers were affiliated with Waymo. At the time of that study, Los Angeles also represented only about 6.45 million of the 56.7 million miles, and the city-specific comparisons had not yet accumulated enough exposure for statistically significant conclusions. Older research involving Swiss Re similarly found fewer liability claims associated with Waymo than calibrated human-driver benchmarks, but that research also involved collaboration with Waymo. The evidence is encouraging and increasingly substantial, but independent analysis remains valuable as robotaxis expand into more environments and accumulate much larger datasets.
Serious Incidents Still Demand Scrutiny
Percentages can make rare events look abstract until one involves a child. On January 23, a driverless Waymo Jaguar I-Pace struck a 9-year-old student who entered the road between stopped vehicles near a Santa Monica elementary school. The National Transportation Safety Board says the Waymo had been traveling at 17 mph before braking. The child suffered minor injuries and did not require transportation for medical treatment. The NTSB investigation remains open, and NHTSA also opened a preliminary evaluation.
Another July collision near downtown Los Angeles sent two people to hospitals and left at least one other person requiring treatment at the scene. According to Waymo’s account, a sedan traveling substantially faster than the robotaxi first struck a pedestrian and then crossed into the Waymo’s path, although police were still investigating the circumstances when the incident was reported. Examples like these explain why aggregate safety advantages do not end the debate. Even if a technology reduces overall risk, regulators still need to understand individual failure modes and determine whether preventable hazards remain.
Los Angeles Is Becoming a Test of What Happens at Real Scale
Robotaxi safety is moving into a different phase. The question is no longer whether a handful of prototypes can drive around selected neighborhoods without crashing. Commercial fleets are carrying paying passengers, accumulating tens of millions of miles and interacting every day with delivery trucks, emergency vehicles, cyclists, pedestrians and unpredictable human drivers. UC Berkeley’s crash dashboard was created specifically to make that growing body of government data easier for researchers and the public to examine.
The 186 Los Angeles crashes should therefore be neither dismissed nor treated as proof that autonomy has failed. The most meaningful evidence points in two directions at once: Waymos are appearing in more reported collisions as their mileage grows, yet most of those collisions cause no apparent injuries and the crash rate per mile has been falling. Current research also suggests meaningful reductions in injury-producing crashes compared with human benchmarks. The unanswered question is how well that performance holds as robotaxis spread beyond carefully defined operating environments. Los Angeles is helping produce the data that may finally provide an answer.
