NetGoodIndexSubmit a correction

Benefit Ledger · verified · Engineering

Waymo publishes lower police-reported crash rates than human benchmarks in rider-only service

Waymo reported that its rider-only Driver logged millions of miles with fewer injury and police-reported crashes than adjusted human baselines in the cities where it operates, in company safety studies released 2023–2024.

20 Dec 2023Tier 3 MajorMethodology 0.1

Current score

+2.16

10 base · Major (tier 3 of 5, 10 pts)
× 0.7000 attribution · Critical contribution
× 0.7000 evidence · Peer review or independent validation
× 0.8000 realization · Independently validated or deployed
× 0.5500 durability
Event-level product before credit split: 2.16

Real-world injury-crash reduction in limited cities is major for AV safety (tier 3). Attribution is to the full AV stack, not a public LLM. Peer-reviewed comparison plus NHTSA SGO inputs (0.70). Realization is paid rider-only service. Durability depends on continued operations.

What happened

Comparisons used matched human-crash estimates from state and insurance data. Independent researchers have generally treated the direction of the result as credible while debating residual selection effects and exposure definitions. The stack is a specialized AV system, not a general LLM. Service remains geographically limited.

Model attribution

+2.16

Waymo Driver

Autonomous driving system operating rider-only vehicles.

Crash outcomes are attributed to the deployed Driver stack.

Attribution 0.7000 · Credit share 100% · Waymo

Claims

  • Waymo reported lower police-reported and injury crash rates than comparable human benchmarks over multi-million-mile rider-only operations.

    outcome · supported

Sources

primary sources

Secondary domains: Medicine & Health

Revision history

  • 13 Sep 2026 · 0.00 → 2.16

    Initial adjudicated seed score under methodology 0.1.