IIHS Hands Waymo an Independent Safety Number — and the Industry a Reporting Problem

The Insurance Institute for Highway Safety has done something the autonomous vehicle sector has lacked since robotaxis began appearing on American streets: produced a safety comparison built by an organisation with no commercial stake in the outcome. Its analysis covers 50 million miles of Waymo operation between 2021 and 2024, set against 222 billion miles logged by human drivers over the same period, and it finds Waymo's driverless fleet crashed 68% less often overall, with single-vehicle crashes down 85% and injury crashes down 81%. For an industry that has spent years fielding safety claims sourced almost entirely from AV operators themselves, that is a meaningful shift in who gets to say the numbers are real.

The scale of the human comparison set is worth sitting with. Waymo's 50 million miles represent a rounding error against 222 billion human miles, and the IIHS is candid that its findings apply to limited, geofenced operation rather than a proof that driverless technology is broadly safer everywhere, on every road, in every condition. IIHS President David Harkey's framing was measured rather than triumphant: driverless systems avoid the impairment, drowsiness and attention lapses that account for a meaningful share of human error, but the data infrastructure to monitor this safely at scale simply does not exist yet. That caveat should carry as much weight for fleet and risk executives as the headline percentage.

The more consequential detail sits in the methodology rather than the topline. IIHS researchers had to strip out crashes that occurred while a human safety operator was present, then apply the police-reporting threshold that would have applied had a human been driving, because driverless operators are obligated to report nearly every incident — including the sort of minor kerb strike that a human driver would quietly forget by the next set of traffic lights. The IIHS estimates roughly half of all property-damage crashes and a third of injury crashes involving human drivers go unreported altogether. That asymmetry means every AV safety comparison to date has effectively been measuring near-total disclosure against a baseline riddled with gaps, and it is a genuine methodological headache rather than a footnote.

For automotive leaders, the practical takeaway is less about Waymo's specific percentage and more about what an independent benchmark unlocks. Insurers pricing AV risk, regulators drafting reporting standards, and OEMs weighing partnerships or their own driverless ambitions have all been operating on incomplete or self-reported data. An IIHS-validated methodology gives the sector a template that could eventually be applied to competitors, and it arrives at a moment when Waymo is expanding into new metros while also absorbing recalls and federal scrutiny — the kind of dual narrative that makes an outside referee genuinely useful to the company's public position.

The timing sits inside a wider regulatory conversation that leaders should track closely. NHTSA has separately proposed removing the manual brake pedal mandate for vehicles designed to operate without a human driver, a change that would reshape robotaxi hardware requirements just as this safety data starts circulating. If crash-reporting standards get harmonised across human and driverless fleets in response to the IIHS's stated concerns, the resulting framework will shape how every OEM with AV ambitions — not just Waymo — gets to make its safety case in public. Given how selectively human drivers currently self-report their own kerb strikes, a tighter standard is arguably overdue for both sides of the comparison.