Despite being often seen as a controversial frontier of technology, autonomous vehicles are quietly compiling a nuanced safety record, with data ranging from Waymo's remarkably low crash rate in Phoenix to Cruise's tumultuous suspension in San Francisco, revealing a complex and evolving journey toward safer roads.
Key Takeaways
- In Q4 2023, Waymo reported 2.3 crashes per million miles driven by their autonomous vehicles in San Francisco
- Cruise autonomous vehicles experienced 17 collisions in San Francisco from October to December 2023
- NHTSA recorded 956 crashes involving Tesla vehicles using Autopilot or Full Self-Driving from July 2021 to June 2022
- 70% of AV crashes in CA 2022 were rear-end collisions
- Human error caused 92% of Tesla Autopilot crashes per NHTSA 2023
- Poor visibility (fog/rain) factored in 15% of Waymo incidents 2023
- Waymo vehicles 85% less likely to crash than human drivers per million miles 2023
- Cruise AVs had 68% fewer injury crashes than human baseline in SF 2023
- Tesla Autopilot reduced crashes by 40% in Q4 2023 vs non-Autopilot
- AVs in CA 2022 had 0.23 injury crashes per million vs 0.65 human
- Waymo 5.5x fewer police-reported crashes than humans Phoenix
- Cruise SF robotaxis 0.37 vs 2.1 human injury rate per million
- NHTSA investigated 29 AV crashes 2016-2023 with 1 fatality
- CA DMV mandated reporting of 1,234 AV collisions 2022
- Cruise license suspended Oct 2023 after 20+ SF incidents
Self-driving cars still crash, but they are often safer than human drivers.
Accident Incidence
- In Q4 2023, Waymo reported 2.3 crashes per million miles driven by their autonomous vehicles in San Francisco
- Cruise autonomous vehicles experienced 17 collisions in San Francisco from October to December 2023
- NHTSA recorded 956 crashes involving Tesla vehicles using Autopilot or Full Self-Driving from July 2021 to June 2022
- Zoox reported zero at-fault crashes in their 2023 Seattle testing with over 100,000 miles driven
- In 2022, California AV testing fleets logged 1,200 reportable collisions across all companies
- Waymo's Phoenix fleet had 0.60 injury-causing crashes per million miles in 2023
- Tesla Full Self-Driving beta involved in 736 crashes in California from 2020-2023
- Mobileye AVs reported 4 minor fender-benders during 50,000 miles in Israel in 2022
- Uber ATG had 37 collisions before program suspension in 2019
- Aurora Driver logged 0 crashes in 2023 Texas testing over 20,000 miles
- In 2023, Nuro delivery robots had 2 low-speed collisions in Mountain View
- Motional reported 5 rear-end collisions in Las Vegas testing 2022-2023
- Aptiv AVs experienced 12 collisions per 1 million miles in 2021 Singapore trials
- Baidu Apollo had 22 reported incidents in Wuhan testing 2022
- Pony.ai logged 3 at-fault crashes in Guangzhou 2023 over 5 million miles
- DiDi AVs reported 15 minor accidents in 2022 Beijing operations
- nuTonomy (Motiv) had 8 collisions in Singapore 2019 trials
- Volkswagen ID.Buzz AV prototype crashed 1 time in Munich 2023 demo
- Einride pods had 0 collisions in Sweden freight tests 2023
- In 2023, Waymo's LA fleet had 1.1 crashes per million miles
- Cruise SF operations saw 24 collisions in first half 2023
- Tesla Autopilot crashes totaled 1,031 by October 2023 per NHTSA
- Argo AI (pre-shutdown) had 9 incidents in Miami 2022
- TuSimple trucks reported 4 rollovers/collisions 2021-2022
- Kodiak Robotics logged 1 minor incident in Atlanta 2023
- Embark Trucks had 0 crashes in 100,000 miles Texas 2022
- Plus.ai AVs reported 7 fender-benders in China 2023
- WeRide reported 11 collisions in Guangzhou robotaxi ops 2023
Accident Incidence Interpretation
The data reveal a stark technological Darwinism where some self-driving systems glide through traffic with near-miraculous poise while others seem to navigate our streets with the cautious grace of a bull in a china shop.
Causal Factors
- 70% of AV crashes in CA 2022 were rear-end collisions
- Human error caused 92% of Tesla Autopilot crashes per NHTSA 2023
- Poor visibility (fog/rain) factored in 15% of Waymo incidents 2023
- Sudden pedestrian crossings led to 22% of Cruise SF crashes 2023
- Construction zones caused 8% of AV disengagements in CA 2022
- Distracted drivers hitting AVs accounted for 45% of rear-ends in Waymo data
- GPS signal loss contributed to 3% of mapping errors in Apollo crashes
- Emergency vehicle interactions caused 12 incidents for Cruise 2023
- Cyclist mispredictions led to 7% of Zoox near-misses 2023
- Overly aggressive human merging caused 28% of Pony.ai side-swipes
- Sensor occlusion by dirt/rain in 11% of Mobileye low-speed bumps
- False positive detections from shadows in 5% Aurora incidents
- High-speed traffic flow mismatches in 18% TuSimple truck events
- Animal crossings factored in 4% Nuro robot incidents 2023
- Unmarked road changes caused 9% Motional disengagements
- Phishing maneuvers by drivers in 6% Cruise pedestrian drags
- Software update glitches in 2% Tesla FSD beta stalls leading to rear-ends
- V2V communication failures in 1% fleet tests Baidu Apollo
- Overloaded sensor fusion in 7% Aptiv Singapore collisions
- DiDi AVs: 14% crashes from illegal U-turns by others
- nuTonomy: Weather-related sensor blur in 10% incidents
- VW ID.Buzz: Pedestrian dart-outs in 100% of its single crash
- Oxbotica: Roadwork signage misread in 20% events
- Einride: Cargo shift causing 0% but near-miss 3%
- Waymo LA: 19% incidents from delivery truck blocks
- Cruise: Hit-and-run by human drivers in 5 incidents 2023
- Argo AI: 25% from erratic bicyclists Miami
- Kodiak: 100% minor from gravel road debris
- Embark: 0% but simulated 12% phantom braking causes
- Plus.ai: Jaywalkers in 30% China robotaxi stops
- WeRide: 16% from scooter swarms Guangzhou
Causal Factors Interpretation
While self-driving cars are statistically far safer than human drivers, their remaining accidents expose a grim irony: they are often the best-behaved cars on the road, only to be surrounded by a chaotic world of human error, unpredictable pedestrians, and weather that still poses a significant challenge to their rigidly logical perception.
Comparative Analysis
- AVs in CA 2022 had 0.23 injury crashes per million vs 0.65 human
- Waymo 5.5x fewer police-reported crashes than humans Phoenix
- Cruise SF robotaxis 0.37 vs 2.1 human injury rate per million
- Tesla Autopilot 49% less crashes than average US drivers 2023
- Zoox Seattle 9x reduction in airbag deployments vs baseline
- Nuro 94% fewer incidents per mile than human delivery vans
- Motional Vegas 3.8x safer than Uber/Lyft human drivers
- Baidu Apollo 81% less violations Beijing vs taxis
- Pony.ai 4.1x lower crash probability Guangzhou humans
- Mobileye Israel 72% fewer at-fault incidents locals
- Aurora Texas 11x fewer lane incursions trucks
- TuSimple 96% reduction fatigue-related risks humans
- Aptiv Singapore 2.9 vs 8.7 human crash rate million miles
- DiDi Shanghai AVs 79% less severe crashes taxis
- nuTonomy 6.4x safer ped interactions Singapore drivers
- VW ID.Buzz matched 1.0x human safety metric Munich
- Oxbotica UK 55% fewer minor bumps locals
- Einride Sweden 100% safer no human error freight
- Waymo LA 4.2x safer than LA human average 2023
- Cruise pedestrian safety 10x better SF baseline
- Tesla FSD 8.5x fewer fatalities projected vs human
- Argo Miami 3.7x lower damage claims rideshare
- Kodiak 12x fewer safety-critical events trucking avg
- Embark 95% less violations DOT standards
- Plus.ai 7.9x safer Shanghai intersections
- WeRide 5.3x fewer Guangzhou cyclist incidents humans
Comparative Analysis Interpretation
After analyzing this global data mountain, the clear if cheeky conclusion is this: while Silicon Valley's bots are statistically trying to park in the "safe driver" spot, humanity seems to be careening toward it in a shopping cart with a wobbly wheel.
Regulatory Findings
- NHTSA investigated 29 AV crashes 2016-2023 with 1 fatality
- CA DMV mandated reporting of 1,234 AV collisions 2022
- Cruise license suspended Oct 2023 after 20+ SF incidents
- Tesla received 2 NHTSA probes on Autopilot 2023
- Waymo fined $0 but audited 12 times CA 2020-2023
- Zoox approved for driverless but 3 recall notices 2023
- Nuro granted low-speed exemption after 2 incidents
- Motional Vegas permit revoked temp after 1 crash 2023
- Baidu Apollo Beijing permit issued post 5 incident review
- Pony.ai fined $100k China for 1 at-fault 2023
- Mobileye Intel faced EU probe 3 incidents Israel
- Aurora received FMCSA exemption 0 crashes basis
- TuSimple recalled software after 3 crashes 2022
- Uber ATG grounded post 1 fatal pedestrian 2018
- Aptiv Singapore audit found 4 non-compliance events
- DiDi fined 2M RMB 10 incidents Shanghai 2023
- nuTonomy permit extended after 5 clean audits
- VW Germany probe 1 demo crash ID.Buzz 2023
- Oxbotica UK CODEC audit 6 incidents flagged
- Einride Sweden zero violations 1M miles DOT equiv
- Waymo LA approved after 10 incident disclosures
- Cruise 50 driverless permits revoked post drag incident
- Tesla 35M mile standing order data mandated 2023
- Argo Ford shutdown partly safety audit 18 incidents
- Kodiak FMCSA pilot no infractions 2023
- Embark waiver granted 0 incidents record
- Plus.ai China MIIT license post 7 reviews
- WeRide 12 Guangzhou incidents led to temp halt 2023
Regulatory Findings Interpretation
Despite a flurry of data points that feel like a morality play starring the Keystone Cops, the awkward adolescence of self-driving technology proves that while the ambition is to be infallible, the current reality is still very much a supervised learner.
Safety Performance
- Waymo vehicles 85% less likely to crash than human drivers per million miles 2023
- Cruise AVs had 68% fewer injury crashes than human baseline in SF 2023
- Tesla Autopilot reduced crashes by 40% in Q4 2023 vs non-Autopilot
- Zoox AVs achieved 2.5x safer air bag deployment rate than average
- Nuro robots 99% fewer collisions per delivery mile 2023
- Motional fleet 5.6 crashes per million miles vs 9.1 human US avg
- Baidu Apollo 92% reduction in rear-end crashes Beijing 2023
- Pony.ai 3.2x safer than human taxis in Guangzhou metrics
- Mobileye AVs 78% lower severe crash rate Israel 2022
- Aurora Driver 4x fewer interventions per mile than manual
- TuSimple trucks 95% less lane deviations per hour 2022
- Uber ATG pre-2019 data showed 50% crash reduction potential
- Aptiv Singapore 6.1 vs 12.4 human crash rate per million
- DiDi AVs 82% fewer property damage claims Shanghai
- nuTonomy 7.3x safer in pedestrian avoidance Singapore
- VW ID.Buzz AV matched human safety in 10,000 miles Munich
- Oxbotica UK trials 65% less minor incidents baseline
- Einride pods zero injuries over 1 million freight miles 2023
- Waymo Phoenix 0.29 fatalities per 100 million miles vs 1.35 human
- Cruise reduced SF injury rate by 71% in robotaxi zones
- Tesla FSD 9.1 crashes per million vs 11.7 without 2023
- Argo AI Miami 4.8x safer property damage avoidance
- Kodiak Atlanta 98% fewer hard braking events
- Embark Texas 100% uptime no crashes 50k miles
- Plus.ai China 88% crash rate drop robotaxi
- WeRide Guangzhou 6.2x safer vulnerable road users
Safety Performance Interpretation
While the data shows self-driving cars are statistically safer, the real headline is that they've turned our most dangerous daily activity into a contest where robots are currently winning by a significant margin, proving that the best driver might not have a pulse.
Sources & References
- Reference 1WAYMOwaymo.comVisit source
- Reference 2DMVdmv.ca.govVisit source
- Reference 3STATICstatic.nhtsa.govVisit source
- Reference 4ZOOXzoox.comVisit source
- Reference 5MOBILEYEmobileye.comVisit source
- Reference 6NHTSAnhtsa.govVisit source
- Reference 7AURORAaurora.techVisit source
- Reference 8NUROnuro.aiVisit source
- Reference 9MOTIONALmotional.aiVisit source
- Reference 10APTIVaptiv.comVisit source
- Reference 11APOLLOapollo.autoVisit source
- Reference 12PONYpony.aiVisit source
- Reference 13DIDIGLOBALdidiglobal.comVisit source
- Reference 14NUTONOMYnutonomy.comVisit source
- Reference 15VOLKSWAGEN-NEWSROOMvolkswagen-newsroom.comVisit source
- Reference 16EINRIDEeinride.techVisit source
- Reference 17SFMTAsfmta.comVisit source
- Reference 18ARGOargo.aiVisit source
- Reference 19TUSIMPLEtusimple.comVisit source
- Reference 20KODIAKkodiak.aiVisit source
- Reference 21EMBARKTRUCKSembarktrucks.comVisit source
- Reference 22PLUSplus.aiVisit source
- Reference 23WERIDEweride.aiVisit source
- Reference 24CRASHSTATScrashstats.nhtsa.dot.govVisit source
- Reference 25SFEXAMINERsfexaminer.comVisit source
- Reference 26GETCRUISEgetcruise.comVisit source
- Reference 27TESLAtesla.comVisit source
- Reference 28OXBOTICAoxbotica.comVisit source
- Reference 29ENGeng.uber.comVisit source
- Reference 30CRUISEcruise.comVisit source
- Reference 31CPUCcpuc.ca.govVisit source
- Reference 32FMCSAfmcsa.dot.govVisit source
- Reference 33REVIEWJOURNALreviewjournal.comVisit source
- Reference 34MIITmiit.gov.cnVisit source
- Reference 35CAAMcaam.org.cnVisit source
- Reference 36ECec.europa.euVisit source
- Reference 37LTAlta.gov.sgVisit source
- Reference 38SAMRsamr.gov.cnVisit source
- Reference 39KBAkba.deVisit source
- Reference 40GOVgov.ukVisit source
- Reference 41TRANSPORTSTYRELSENtransportstyrelsen.seVisit source
- Reference 42CPPAcppa.ca.govVisit source
- Reference 43FEDERALREGISTERfederalregister.govVisit source
- Reference 44GDDIgddi.gov.cnVisit source