What are the odds of being killed by a self-driving car?
Evidence quality 4.38/5
Eight-dimension review score against the quality rubric . Each dimension scored 1–5.
- D1 Source grounding
- 5/5
- D2 Source authority
- 5/5
- D3 Arithmetic
- 4/5
- D4 Uncertainty
- 4/5
- D5 Scope
- 3/5
- D6 Prose
- 5/5
- D7 Perception honesty
- 4/5
- D8 Caveat completeness
- 5/5
● your factors — click this risk ▾ to reveal
- Your factors
≈ As likely as
Perceived
Autonomous vehicles occupy an outsized share of public fear relative to their actual deployment. Gallup and AAA surveys consistently find that roughly three in four Americans say they would be afraid to ride in a fully self-driving car. The 2018 Uber ATG fatality in Tempe, Arizona — the first pedestrian killed by a vehicle operating in autonomous mode — anchored public perception in a way that tens of thousands of annual human-driver fatalities do not. Every subsequent AV incident receives national coverage; the 39,000-plus human-driver deaths per year are reported as a quarterly statistical abstract.
Rough estimate: most Americans rate self-driving cars as more dangerous than human drivers
Source: editorial intuition, not polled
Actual
~2 ADS fatalities across ~170 million autonomous miles (US, through 2025)
US fully autonomous (Level 4-5) vehicles
Show derivation
As of late 2025, NHTSA's Standing General Order data records 2 fatalities involving vehicles operating under a fully Automated Driving System (ADS, Levels 4-5). Crucially, neither fatality was caused by the ADS itself — both involved other at-fault human drivers striking the ADS vehicle. The at-fault ADS fatality rate is therefore 0. However, the "involved-in" rate still matters: it measures the risk of dying in or near an ADS vehicle regardless of fault. The Waymo fleet — the largest ADS operator — has driven 170+ million fully autonomous miles with zero at-fault fatalities. To normalize the involved-in rate: 2 fatalities / 170,000,000 miles ≈ 1.18 per 100 million miles. A US adult drives roughly 740,000 miles over a 59-year adult lifetime (12,500 mi/yr). If all of those miles were driven by ADS at the observed involved-in rate, the lifetime risk would be 2/170,000,000 × 740,000 ≈ 0.0087 (~1 in 115). This is comparable to the human-driver lifetime fatality risk (~0.0095). The uncertainty band is wide: the low end reflects the at-fault ADS rate (effectively 0; a structural floor of 0.000001 is used), while the high end accounts for small-sample volatility. This figure is inherently speculative because ADS deployment remains limited to select urban geofences. The 2026-06 review added a peer-reviewed source (Scanlon et al., Traffic Injury Prevention, 56.7M rider-only miles through Jan 2025) that reports injury-crash reductions, not a fatality count — it does not move the point estimate but reinforces the low end of the band, since the zero-at-fault-fatality record now spans 56.7M+ driverless miles.
Caveats: This headline measures the fatal outcome for a person in or near a fully driverl…
This headline measures the fatal outcome for a person in or near a fully driverless (Level 4-5) ADS vehicle, and the 2-death numerator is too small for confidence. Several 2024-2026 developments sit outside that frame and are not folded into the number. (1) The victim and severity profile is shifting: AV-as-striker-of-a-vulnerable-pedestrian incidents (e.g., a Jan 2026 Waymo child strike near a Santa Monica school, plus a Dec 2025 Waymo recall over failure to stop for school buses loading/unloading) tend to be non-fatal but are a different harm mode than the rider/at-fault-driver framing here; clean lifetime figures for them lean on small recall/incident dockets. (2) The named fatal failures in the public record are almost all Level-2 driver-assist (Tesla Autopilot/FSD), not full ADS — captured by the Level-2 monitoring-disengagement multiplier, not this rate. NHTSA's reduced-visibility FSD probe (PE24031) and the Oct 2025 traffic-law-violation probe (PE25012, 2.88M vehicles, 14 crashes / 23 injuries) document Level-2 failure modes (camera degradation in glare/fog/dust, running red lights) without a per-mile fatality denominator. (3) Automation-induced drowsiness and complacency while supervising Level-2 assist is documented by 2025 systematic reviews to worsen takeover performance, but rests on simulator/behavioral findings with no crash-rate denominator, so no estimate is offered for it.
Related risks
Other risks on similar themes — for exploring related fears.
Driving after cannabis
What are the odds of causing a fatal crash by driving within a few hours of using cannabis?
Driving on sedating meds
What are the odds of causing a fatal crash by driving while on a sedating prescription medication?
Hand-held phone call + driving
What are the odds of a crash from holding a phone to your ear while driving?
Pick challenger
Across roughly 170 million fully autonomous miles driven in the United States through late 2025, NHTSA’s Standing General Order database records 2 fatalities involving vehicles operating under a Level 4 or 5 Automated Driving System. Neither was caused by the ADS itself — both involved other human drivers striking the autonomous vehicle. The at-fault ADS fatality count is zero. The involved-in fatality rate, about 1.2 per 100 million miles, is in the same neighborhood as the human-driver rate of 1.2-1.3 per 100 million miles, but the ADS sample is vanishingly small: two deaths versus 39,000. Whether to count involved-in or at-fault matters enormously — one framing gives a lifetime risk around 1 in 115, the other gives effectively zero. Waymo, which accounts for the bulk of those autonomous miles, reports an 85 percent reduction in injury-causing crashes compared to human benchmarks, a figure independently corroborated by Swiss Re insurance claims data.
The fear runs far ahead of the deployment. Three in four Americans tell pollsters they would be afraid to ride in a self-driving car, yet the number of Americans who have actually shared a road with an ADS vehicle in any given week is still a rounding error. The Uber ATG pedestrian fatality in 2018 — a single death under circumstances that involved a distracted safety driver and a poorly configured system — did more to shape public risk perception than the subsequent 150-plus million uneventful autonomous miles that followed. This is textbook availability bias: one vivid, novel event anchors a probability estimate that thousands of data points cannot dislodge.
Where all of this falls apart as a predictive tool: ADS vehicles currently operate in curated urban geofences with favorable weather, well-mapped roads, and low-speed limits. Whether the observed safety record scales to highway driving, rural roads, snow, or a fleet a thousand times larger is genuinely unknown. The 2-fatality numerator is also too small to draw statistical conclusions with any confidence — the uncertainty band spans more than two orders of magnitude. The number published here is what the data say so far, not a forecast.
Related tidbits
If every mile you drove were on autonomous systems, the lifetime odds of being killed by one come out around ~1 in 115 lifetime, built from about 2 ADS fatalities across roughly 170 million autonomous miles logged through 2025.
Claim ledger
Every number below is what each source reported, with the verbatim quote we relied on and how we arrived at our figure. Click any link to verify directly.
-
[1] National Highway Traffic Safety Administration (NHTSA) — Standing General Order on Crash Reporting for Incidents Involving ADS and Level 2 ADAS
Standing General Order on Crash Reporting for Incidents Involving ADS and Level 2 ADAS- Statistic
2 fatalities involving vehicles operating under ADS (Levels 3-5) through November 2025; 54 fatalities involving Level 2 ADAS- Excerpt
“"Fully self-driving car deaths were equal to only 2 as of the reporting period. Cars with ADAS Level 2, which are not fully automated, have been involved in 54 deaths." ”
- Source data from
- 2025-11-17
- Accessed
- 2026-04-18 · archived copy
- Calculation
- NHTSA's Standing General Order mandates crash reporting for all ADS and Level 2 ADAS vehicles. The 2 ADS fatalities are the numerator (involved-in, not at-fault — neither was caused by the ADS). Denominator is estimated total ADS miles driven across all operators (Waymo, Cruise, Zoox, etc.), with Waymo's 170M+ miles forming the bulk. Rate ≈ 2 / 170,000,000 ≈ 1.18 per 100 million miles. Lifetime = rate × 740,000 lifetime miles ≈ 0.0087 (~1 in 115).
-
[2] Waymo LLC (Alphabet) — Waymo Safety Impact
Waymo Safety Impact- Statistic
85% reduction in injury-causing crashes vs human benchmark over 170+ million autonomous miles; zero at-fault fatalities- Excerpt
“"The Waymo Driver demonstrated an 85% reduction or 6.8 times lower crash rate involving any injury, from minor to severe and fatal cases (0.41 incidence per million miles for the Waymo Driver vs 2.78 for the human benchmark)." ”
- Source data from
- 2025-12-01
- Accessed
- 2026-04-18 · archived copy
- Calculation
- Waymo's self-reported safety data covers rider-only (fully driverless) operations. The 85% injury-crash reduction is corroborated by Swiss Re insurance claims data (92% fewer bodily injury claims over 25 million miles). Used here to contextualize the per-mile safety performance rather than as the primary rate source.
- Independence
- Waymo's safety data is self-reported but has been independently validated by Swiss Re (insurance claims analysis) and peer-reviewed in the journal Traffic Injury Prevention (Scanlon et al., 2024). The NHTSA SGO data above is the independent government source.
-
[3] Traffic Injury Prevention (Scanlon et al.) — Comparison of Waymo Rider-Only crash rates by crash type to human benchmarks at 56.7 million miles
Comparison of Waymo Rider-Only crash rates by crash type to human benchmarks at 56.7 million miles- Statistic
85% fewer suspected-serious-injury+ crashes and 92% fewer pedestrian-injury crashes across 56.7 million rider-only miles through end of January 2025; zero at-fault fatalities- Excerpt
“"The Waymo Driver had statistically significant lower crashed vehicle rates than the human benchmarks for any-injury-reported crashes, airbag deployment crashes, and suspected serious injury+ crashes... an 85% reduction (95% CI 39% to 99%) in Suspected Serious Injury+ crashes... with substantial reductions in crashes involving injuries among pedestrians (92% reduction) and cyclists (82% reduction)." ”
- Source data from
- 2025-05-01
- Accessed
- 2026-06-14
- Calculation
- Peer-reviewed retrospective safety assessment (accepted in Traffic Injury Prevention, 2025) of Waymo rider-only operations through end of January 2025. It does not change the ADS fatality numerator (this entry's headline is a fatality figure; the study reports injury-crash rates), but it independently corroborates the zero-at-fault-fatality record across 56.7M+ driverless miles, reinforcing the low end of the uncertainty band and backing the geofenced-route protective multiplier. Used as authoritative contextualization, not as the primary rate source.
-
[4] National Highway Traffic Safety Administration (NHTSA) — Early Estimate of Motor Vehicle Traffic Fatalities for 2024
Early Estimate of Motor Vehicle Traffic Fatalities for 2024- Statistic
39,254 people killed in motor vehicle traffic crashes in 2024- Excerpt
“"An estimated 39,254 people died in motor vehicle traffic crashes in 2024." ”
- Source data from
- 2025-04-01
- Accessed
- 2026-04-18 · archived copy
- Calculation
- The 39,254 human-driver-era deaths provide the baseline for comparison. At ~3.2 trillion miles driven annually, the human fatality rate is ~1.23 per 100 million miles in 2024, roughly comparable to the ADS observed rate but based on a sample size billions of times larger.