What does Waymo's 2025-2026 safety report show, and is Waymo safer than human drivers?

What Waymo publishes

Waymo publishes annual safety reports as part of its operational disclosure. The reports document:

• Cumulative rider-only autonomous miles operating: the verified mileage base across Waymo's commercial robotaxi service.
• Police-reported crashes per million miles: compared to human-driver baselines from NHTSA's general traffic-safety data and academic studies of comparable urban environments.
• Injury-causing crashes per million miles: separated from total crashes because the severity distribution matters editorially.
• Pedestrian and cyclist incidents per million miles: a specific subset given the disproportionate vulnerability of those road users.
• Published methodology: how Waymo classifies incidents (severity, contributing factors, root cause), what counts as an incident, and how the comparison to human-driver baselines is constructed.

The reports also include third-party validation from Swiss Re, the global reinsurance company. Swiss Re's actuarial analysis of Waymo's safety data uses insurance-industry methods to evaluate per-incident frequency relative to comparable human-driver insurance baselines.

What the 2025-2026 data shows

Per Waymo's published safety reporting through 2025-2026:

• Police-reported crash rate per million miles: substantively below human-driver baselines for comparable urban environments. The favorable differential is documented at meaningful effect-size, not at the noise floor.
• Injury-causing crash rate per million miles: similarly favorable to Waymo, with the differential more pronounced at higher severity tiers.
• Pedestrian and cyclist incidents per million miles: favorable to Waymo in absolute frequency, though sample sizes for the rarest severe-injury incidents involve small numbers that make ratio comparisons methodologically sensitive.
• Swiss Re actuarial validation: confirms the per-incident frequency falls below comparable human-driver insurance baselines using standard actuarial methods.

The framework's verified posture on Waymo safety is: data depth verified through multi-year operational history, third-party actuarial validation, regulator-required NHTSA reporting, and published methodology that academic and industry analysts can review.

The methodological complexity

Per DEPLOY's verified-vs-claimed framework on safety incidents and recalls, favorable Waymo comparative statistics should be cited with methodological context. Three nuances that any honest reading should preserve:

• Urban versus suburban operating context: Waymo operates primarily in dense urban contexts with relatively predictable traffic patterns, well-maintained infrastructure, and modest speed envelopes. Human-driver-baseline data spans all driving contexts including high-speed rural highways, weather-extreme conditions, and infrastructure-variable environments. Apples-to-apples comparison requires controlling for operating context.
• Sample size and incident rarity: severe injury and fatality incidents are rare events. Statistical comparison at multi-year multi-city scale produces meaningful per-million-miles rates; per-city or per-quarter comparisons can be noise-limited.
• Reporting-standard differences: Waymo's per-incident reporting standard is detailed (every minor contact, every police interaction). Human-driver baseline studies vary; some studies under-count minor incidents because they go unreported. This can produce false-favorable comparisons if not controlled for.

The methodological honesty does not invert the favorable comparison; it qualifies the comparison. Waymo's safety record is materially better than human-driver baselines on the dimensions documented; the precise effect size depends on methodological choices that any honest reading should preserve.

Is Waymo safer than human drivers?

The framework's honest answer: yes, on the dimensions Waymo's reporting documents, controlled for the comparison contexts that methodologically apply. Specifically:

• In comparable urban environments: Waymo's police-reported and injury-causing crash rates are substantively below human-driver baselines.
• At multi-year multi-city operational scale: the favorable comparison is robust across geographies and time periods, not a single-context artifact.
• Per Swiss Re actuarial validation: the insurance-industry analysis confirms the comparison favorable on standard actuarial methods.

What the framework does not claim:

• That Waymo's safety record extends to non-Waymo operating contexts (rural highways, severe weather, infrastructure-variable environments where Waymo does not operate).
• That every per-million-mile statistic generalizes to every consumer query about safety. Specific safety questions ("is it safe to ride at night in heavy rain") may have different answers than the aggregate per-million-miles rate.
• That Waymo has eliminated safety incidents. Crashes occur; Waymo's published rate of crashes per mile is meaningfully below human-driver comparable baselines, not zero.

How Waymo's published methodology differs from Tesla's

Across the US robotaxi cluster, published safety methodology depth varies materially:

• Waymo: annual safety reports with per-million-miles metrics + third-party actuarial validation + peer-reviewable methodology + Swiss Re partnership.
• Tesla Robotaxi: pilot-stage incident tracking; broader Tesla-fleet safety data inherited from Autopilot/FSD context; per-pilot published safety analysis lighter than Waymo's annual reports. See how safe is Tesla Robotaxi.
• Tesla overall (multiple products): substantial Autopilot fatality data via NHTSA Standing General Order. See Tesla fatality rate explained for the cross-product disambiguation.

The methodological-depth differential is editorially substantial. Per DEPLOY's vvc-sharper-across-competitive-set discipline, Waymo's verified methodology depth distinguishes it from peer operators at the safety-data layer.

Recent operational events worth noting

Waymo's recent operational record includes:

• The 3,791-vehicle flooding recall (May 2026 NHTSA acknowledgment) anchors the federal-recall-layer safety record. Recall is a remedial action under the federal framework; the underlying defect is documented; the remedy process is regulator-tracked.
• Service pauses in Atlanta and San Antonio (flooding-related) and freeway service halts across four markets (construction-zone perception failures) are documented operational responses that inform the safety-evidence layer alongside the per-million-miles aggregate.

Per the framework, recall actions and operational suspensions are editorially distinct from per-million-miles statistics; both surfaces inform the complete safety reading.

Where to go for context

For the bilateral safety comparison with Tesla Robotaxi, see Waymo vs Tesla Robotaxi safety comparison. For the broader operational comparison, see how Tesla Robotaxi compares to Waymo. For Waymo's fatal-crash history specifically, see how many fatal crashes Waymo has had.

For service-evaluation context: Waymo on the main consumer surface, Waymo's canonical registry record, and the framework DEPLOY applies to safety claims across operators at how DEPLOY tracks safety incidents and recalls. For methodology canonical references applicable to Waymo safety report framing: the 9-tier source-quality rubric (Waymo published safety data + Swiss Re actuarial + NHTSA SGO primary-government-record source classification).