Google's self-driving car head says after 1.7M miles, just 11 minor accidents have resulted, none Google's fault
The View from the Front Seat of the Google Self-Driving Car — After 1.7 million miles we've learned a lot—not just about our system but how humans drive, too.
Context & Ripple Effects
This post is Google's first full self-accounting of its driverless program: 1.7 million miles, 11 minor accidents, zero attributed to its own system. It lands three days before the company puts its own cars on public roads with safety drivers and a 25mph cap, making the safety record the argument for why that deployment is safe.
The disclosure also pre-empts scrutiny that arrived within weeks — the California DMV's release of prototype incident reports confirmed the same pattern of accidents caused by other drivers — and set up the transparency playbook Google formalized a month later with a dedicated accident-data website. The record itself was about to get more complicated.
First-order effects
- Google hands California regulators and the public a clean-slate baseline — no at-fault crashes — immediately before deploying vehicles without test-track conditions, converting the mileage number into regulatory capital.
- Every future incident now gets measured against this 'none Google's fault' framing, which the DMV's subsequent report releases put on official record rather than leaving to company blog posts.
Second-order effects
- Transparency becomes a competitive and PR instrument: once Google publishes per-mile accident counts and builds a site around them, any rival seeking permits or public trust has to match the disclosure standard or look evasive by comparison.
- The framing invites adversarial auditing of the fine print — and indeed later disclosures of 13 near-misses and 272 human interventions over 14 months, followed by a 2mph collision with a public bus that may be Google's first at-fault accident, stress-tested exactly whose fault 'fault' is.
Third-order effects
- If the pattern holds, industry safety arguments migrate from raw mileage anecdotes to statistically benchmarked crash-rate comparisons against human drivers — the structure behind Waymo's eventual claim of an 85% reduction in injury-causing crash rates across 7.14M miles.
- Fault attribution becomes the central contested metric of autonomy regulation: who logs it, how interventions count, and whether a regulator or the operator owns the definition of 'at fault.'
The trend: Autonomous-vehicle safety claims are evolving from self-reported milestone counts toward benchmarked, regulator-audited crash statistics — with fault attribution as the metric everyone fights over.