IIHS study of 5K+ crashes: self-driving cars would have prevented only a third, those caused by “sensing and perceiving” errors and driver “incapacitation”
DETROIT (AP) — A new study says that while autonomous vehicle technology has great promise to reduce crashes …
Context & Ripple Effects
IIHS has spent years building the case that driver-assistance branding outpaces reality: its [[a:983687|survey found 53% of Super Cruise and 42% of Autopilot users comfortable treating their cars as fully self-driving]], with system names like 'Autopilot' itself blamed for inflating expectations. This new study is the quantitative counterpart — across 5,000+ real crashes, it puts a hard ceiling on what full automation would actually have prevented: roughly one-third.
That framing cuts directly against vendor benchmarks like Waymo's claimed 85% reduction in injury-causing crash rates over 7.14M miles, and it lands amid an unresolved dispute over evidence: experts have warned that [[a:979687|the lack of publicly available data makes safety claims by Tesla and others difficult to verify]]. The study matters because it gives insurers and regulators an independent denominator for the automation-safety debate.
First-order effects
- Automakers' safety messaging now has a third-party ceiling to contend with: IIHS attributes the preventable share almost entirely to sensing-and-perceiving errors and driver incapacitation, so 'self-driving prevents crashes' claims must be narrowed to those specific failure modes.
Second-order effects
- The finding sharpens scrutiny of driver-assistance naming and handoff behavior — the same overestimation IIHS documented among Super Cruise, Autopilot, and ProPILOT Assist users becomes harder to defend when the underlying crash math shows two-thirds of crashes lie outside automation's reach.
Third-order effects
- If vendors' self-reported mileage benchmarks and independent crash-causation studies keep diverging, the structural fix points toward standardized public incident data — the transparency gap experts flagged in 2022 becomes the battleground for whether automation safety claims can be regulated at all.
The trend: Autonomous-vehicle safety is shifting from utopian crash-elimination promises toward mode-specific, independently verifiable claims as researchers quantify exactly which crashes automation can and cannot prevent.