Are robotaxis safer than human drivers?

Waymo's published data shows a crash rate substantially below the human-driver baseline (~1.3 fatalities per 100M vehicle miles US, urban lower than highway) for comparable urban environments, with particularly large differentials on high-severity crashes. Third-party actuarial analysis (Swiss Re) corroborates the published metrics. The statistical caveat: confidence intervals at current cumulative mileage are wider than a single headline number suggests. Tesla Robotaxi and Zoox pilot-scale datasets are smaller; comparisons against human-driver baselines are statistically less definitive at pilot scale.

Have any robotaxi passengers died?

As of mid-2026, no robotaxi rider fatalities have been reported across the three major US operators tracked here. Waymo reports zero rider fatalities across tens of millions of autonomous miles per the operator's published periodic safety reports and NHTSA Standing General Order tracking. Tesla Robotaxi reports no fatal crashes at the 4-market paid pilot scope (since June 2025). Zoox operates at free-demo pilot scope with no rider fatalities, though a January 2026 SF injury incident remains under regulatory review.

How does NHTSA track robotaxi safety?

NHTSA's Standing General Order on Crash Reporting (SGO 2021-01) requires AV operators to report every crash involving their vehicles, regardless of fault. The dataset includes at-fault crashes, crashes where a human-driven vehicle hit the AV, and minor contact incidents. Severity reporting timelines vary: severe incidents within one day, all crashes within five days. NHTSA's Part 573 framework governs vehicle recalls (Zoox filed two recalls in 2025: December lane-crossing affecting 332 units, April Las Vegas collision affecting ~270 units). The SGO dataset provides an independent regulatory verification surface accessible to the public.

Is Waymo safer than Tesla Robotaxi?

By available data, Waymo has the stronger public safety record. Waymo operates multi-year commercial service across 11 metros with tens of millions of autonomous miles and published safety reports showing rates below human-driver baselines. Tesla Robotaxi operates a 12-month, 4-market paid pilot with no fatal crashes verified at pilot scope but a substantially smaller public safety dataset. Per DEPLOY's asymmetric-data-comparison framework, the comparison is asymmetric because operational scales are asymmetric, not because Tesla Robotaxi has demonstrated safety problems at pilot scale.

What's a robotaxi recall?

A robotaxi recall is a manufacturer-initiated remediation filed under NHTSA's Part 573 framework when a safety defect or non-compliance is identified across a population of vehicles. Zoox filed two notable recalls in 2025: the December 2025 lane-crossing recall (332 robotaxis affected) and the April 2025 Las Vegas collision recall (approximately 270 robotaxis). The recall record is not pejorative; it reflects the operator participating inside the federal recall framework, which is itself a verification surface for safety claims. Per DEPLOY's methodology on safety incidents and recalls, regulator-anchored remediation records carry distinct editorial weight from operator narrative.

Explainers Robotaxis & autonomous vehicles

Are robotaxis safe?

Yes, on average, at the per-operator scales that have accumulated data. Waymo's published safety record shows a crash rate substantially below the human-driver baseline for comparable urban environments across tens of millions of autonomous miles. Tesla Robotaxi and Zoox operate at pilot scale with thinner public safety datasets but no fatal crashes verified. The fair conclusion per DEPLOY's framework: robotaxi safety is verifiable per-operator at the scales each has accumulated, with statistical confidence intervals tied to cumulative mileage.

Waymo rider fatalities

1.3

Human baseline per 100M mi

US operators tracked

NHTSA SGO

Regulatory surface

Per-operator

Verification posture

Mid-2026

Snapshot date

The framework for evaluating robotaxi safety

Asking "are robotaxis safe" requires first naming which operator + at which operational scale + against which baseline. Per DEPLOY's verified-vs-claimed framework applied to safety, four distinctions matter:

Which operator. Waymo, Tesla Robotaxi, and Zoox operate at different verification tiers with different cumulative-mileage depths.
Which operational scale. Multi-metro commercial-deployed (Waymo 11 metros) produces different safety dataset than 4-market paid pilot (Tesla) or 2-market free demo pilot (Zoox).
Which baseline. Human-driver fatality rate (~1.3 per 100M vehicle miles in the US, urban lower than highway) is the natural comparison; cohort-internal comparisons require equivalent operational envelopes.
Which verification surface. Operator-published safety reports + NHTSA Standing General Order + third-party actuarial analysis (Swiss Re for Waymo) produce different verification layers.

Public conversation often collapses these distinctions. The framework asks to preserve them.

What Waymo's safety record shows (verified at commercial scale)

Waymo publishes periodic safety reports covering tens of millions of fully autonomous rider-only miles across its 11-metro commercial footprint. Through mid-2026, Waymo reports zero rider fatalities. Published metrics include:

Per-million-miles crash rate substantially below human-driver baseline for comparable urban environments, with notably lower rates of high-severity crashes specifically.
Third-party actuarial validation: Swiss Re analyzed Waymo's safety record using actuarial methods and confirmed per-incident frequency falls below comparable human-driver insurance baselines.
NHTSA Standing General Order reporting: Waymo participates in SGO 2021-01, producing publicly accessible incident records.

The framework reads Waymo's safety record as verified across three independent surfaces (operator publication + NHTSA SGO + Swiss Re actuarial). For deeper context, see how many fatal crashes Waymo has had and waymo safety report 2025-2026.

What Tesla Robotaxi's safety record shows (pilot-scale data)

Tesla Robotaxi operates a 4-market paid pilot footprint (Austin lead from June 2025, plus Dallas, Houston, and SF Bay Area). The cumulative operational history is approximately 12 months as of mid-2026, materially shorter than Waymo's 5+ years of commercial driverless operation. Published per-mile safety statistics are thinner than Waymo's multi-year operational baseline; no fatal Tesla Robotaxi crashes have been verified at pilot scope. Per DEPLOY's framework on safety, Tesla Robotaxi's safety record is verifiable at the pilot scale that exists, but the depth is not yet at the level of Waymo's published annual reports. See how safe is Tesla Robotaxi and waymo vs Tesla Robotaxi safety for asymmetric-data-comparison framework.

What Zoox's safety record shows (pilot-tier free demo)

Zoox operates at pilot tier in 2026: free public demo service on the Las Vegas Strip and in SF SoMa. Zoox's deployment record includes documented regulatory engagement: the December 2025 lane-crossing recall (332 robotaxis) and the April 2025 Las Vegas collision recall (approximately 270 robotaxis), both under NHTSA's Part 573 framework. A January 2026 SF injury incident remains under regulatory review. The recall record is not pejorative; it reflects Zoox operating inside the federal recall framework, which is the verification surface for safety claims. Operators evaluating Zoox safety should read the recall record alongside the deployment record as the complete verification picture.

The asymmetric-data comparison

The three US robotaxi operators produce structurally different safety datasets because they operate at structurally different scales:

Waymo: Multi-year commercial driverless operation; tens of millions of autonomous miles; third-party actuarial validation; NHTSA SGO depth. The deepest verifiable safety dataset in the cluster.
Tesla Robotaxi: 12-month paid pilot across 4 markets; no fatal crashes at pilot scope; thinner public dataset.
Zoox: Free demo pilot across 2 markets; 2 NHTSA recalls + 1 injury under review; pilot-scale safety record with regulator engagement.

Per DEPLOY's framework on how to track safety incidents and recalls, per-operator safety records should not be averaged into a single "robotaxi safety" claim because operational scales differ structurally. Each operator's safety verification is anchored at its own per-operator scale.

Comparison to human-driver baseline

In the US, the human-driver fatality rate is roughly 1.3 per 100 million vehicle miles traveled, with urban driving lower than highway driving. Waymo's reported zero rider fatalities across tens of millions of autonomous urban miles is consistent with (or below) the baseline, with the statistical caveat that confidence intervals at current cumulative mileage are wider than a single headline number suggests. The comparison favors robotaxi safety at the per-operator scales that have accumulated data, but the comparison's statistical depth tightens with mileage accumulation.

NHTSA Standing General Order as verification surface

NHTSA's Standing General Order on Crash Reporting (SGO 2021-01) requires AV operators to report every crash involving their vehicles. This dataset includes at-fault crashes, crashes where a human-driven vehicle hit the AV, and minor contact incidents. For Waymo specifically, the great majority of SGO-reported incidents involve a human-driven vehicle hitting a Waymo (often at low speeds in city driving). The distinction between "in a crash" and "caused a crash" is editorially load-bearing. The SGO dataset provides an independent regulatory verification surface across all qualifying AV operators.

Methodological caveats

Self-reporting layer. Operator-published safety reports are the primary public source for per-mile rates. NHTSA's SGO is the regulatory check.
Per-metro variance. Cumulative miles per metro vary substantially (Phoenix has the longest Waymo operational history; newer metros have less data).
Operational-envelope differences. Each operator's safety record applies to its own operational envelope (geofenced areas, time-of-day restrictions, weather conditions). Claims do not generalize across envelopes.
Sample-size confidence intervals. Fatality rate comparisons against human-driver baselines require sufficient mileage to produce statistically definitive claims. The cohort accumulates mileage; confidence tightens over time.

Bottom line

Robotaxi safety is verifiable per-operator at the cumulative scales each has accumulated. Waymo's published record shows a crash rate substantially below the human-driver baseline across tens of millions of autonomous miles, validated by third-party actuarial analysis. Tesla Robotaxi and Zoox operate at pilot scales with thinner public datasets but no fatal crashes verified. The framework reads "are robotaxis safe" as a per-operator question rather than a category-aggregate question, with statistical confidence tied to cumulative mileage at each operator's scale.

For deeper per-operator context, see how many fatal crashes Waymo has had, how safe is Tesla Robotaxi, and which is safer Waymo or Tesla Robotaxi. For DEPLOY's broader safety methodology, see how DEPLOY tracks safety incidents and recalls. For methodology canonical references applicable to robotaxi safety framing: verified-vs-claimed at within-entity granularity (per-operator safety verification at distinct operational scales) + the 9-tier source-quality rubric.

Source: operator-published safety reports + NHTSA Standing General Order + Swiss Re actuarial analysis (Waymo) + Zoox NHTSA recall filings. Tier reflects mid-2026 verified state.
Operator	Rider fatalities	Operational scope	Tier
Waymo	0 reported	11 metros commercial; tens of M autonomous miles	Commercial
Tesla Robotaxi	0 at pilot scope	4-market paid pilot; ~12 months operational	Paid pilot
Zoox	0 at pilot scope	2-market free demo; 2 NHTSA recalls + Jan 2026 incident	Demo pilot
Human-driver baseline	~1.3 per 100M miles (US urban)	National baseline	Baseline

Frequently asked questions

Are robotaxis safe?: Yes, on average, at the per-operator cumulative scales each has accumulated. Waymo's published safety record shows a crash rate substantially below the human-driver baseline (~1.3 per 100M vehicle miles US) across tens of millions of autonomous miles. Tesla Robotaxi and Zoox operate at pilot scales with thinner public safety datasets but no fatal crashes verified. Per DEPLOY's framework, "are robotaxis safe" is a per-operator question; each operator's safety verification is anchored at its own per-operator cumulative scale.
Are robotaxis safer than human drivers?: Waymo's published data shows a crash rate substantially below the human-driver baseline (~1.3 fatalities per 100M vehicle miles US, urban lower than highway) for comparable urban environments, with particularly large differentials on high-severity crashes. Third-party actuarial analysis (Swiss Re) corroborates the published metrics. The statistical caveat: confidence intervals at current cumulative mileage are wider than a single headline number suggests. Tesla Robotaxi and Zoox pilot-scale datasets are smaller; comparisons against human-driver baselines are statistically less definitive at pilot scale.
Have any robotaxi passengers died?: As of mid-2026, no robotaxi rider fatalities have been reported across the three major US operators tracked here. Waymo reports zero rider fatalities across tens of millions of autonomous miles per the operator's published periodic safety reports and NHTSA Standing General Order tracking. Tesla Robotaxi reports no fatal crashes at the 4-market paid pilot scope (since June 2025). Zoox operates at free-demo pilot scope with no rider fatalities, though a January 2026 SF injury incident remains under regulatory review.
How does NHTSA track robotaxi safety?: NHTSA's Standing General Order on Crash Reporting (SGO 2021-01) requires AV operators to report every crash involving their vehicles, regardless of fault. The dataset includes at-fault crashes, crashes where a human-driven vehicle hit the AV, and minor contact incidents. Severity reporting timelines vary: severe incidents within one day, all crashes within five days. NHTSA's Part 573 framework governs vehicle recalls (Zoox filed two recalls in 2025: December lane-crossing affecting 332 units, April Las Vegas collision affecting ~270 units). The SGO dataset provides an independent regulatory verification surface accessible to the public.
Is Waymo safer than Tesla Robotaxi?: By available data, Waymo has the stronger public safety record. Waymo operates multi-year commercial service across 11 metros with tens of millions of autonomous miles and published safety reports showing rates below human-driver baselines. Tesla Robotaxi operates a 12-month, 4-market paid pilot with no fatal crashes verified at pilot scope but a substantially smaller public safety dataset. Per DEPLOY's asymmetric-data-comparison framework, the comparison is asymmetric because operational scales are asymmetric, not because Tesla Robotaxi has demonstrated safety problems at pilot scale.
What's a robotaxi recall?: A robotaxi recall is a manufacturer-initiated remediation filed under NHTSA's Part 573 framework when a safety defect or non-compliance is identified across a population of vehicles. Zoox filed two notable recalls in 2025: the December 2025 lane-crossing recall (332 robotaxis affected) and the April 2025 Las Vegas collision recall (approximately 270 robotaxis). The recall record is not pejorative; it reflects the operator participating inside the federal recall framework, which is itself a verification surface for safety claims. Per DEPLOY's methodology on safety incidents and recalls, regulator-anchored remediation records carry distinct editorial weight from operator narrative.

Robotaxi safety verifies per-operator at cumulative scales. Verification surfaces: operator publications + NHTSA Standing General Order + third-party actuarial analysis (Swiss Re for Waymo). Per-operator data should not be averaged into category-aggregate claims. How DEPLOY verifies →

Compare alternatives

Waymo service overviewTesla Robotaxi serviceWhere Waymo operates

Defined terms in this explainer

More in robotaxis & autonomous vehicles

View all 20 explainers in robotaxis & autonomous vehicles →

← All explainers