Waymo says disengagement rate per 1K miles driven autonomously dropped from 0.8 in 2015 to 0.2 in 2016, even while miles driven rose to 635K overall
Waymo's self-driving software had far fewer disengagements in 2016 than it did in 2015 in the state of California, despite a big increase in the number of miles driven overall.
Context & Ripple Effects
This is Waymo's first full year under California's mandatory disengagement disclosure regime, and the numbers double as a competitive signal: on the same day, Bloomberg reported Tesla ran just four autonomous cars totaling 550 miles in the state, so Waymo is publishing scale and reliability simultaneously. The annual DMV filing becomes the industry's de facto scoreboard — a year later Waymo and GM top it again with 63 and 105 disengagements respectively (the 2018 California DMV report).
The trajectory matters beyond the metric itself: these early test-mile disclosures are the baseline against which Waymo later frames its safety case, from an 85% reduction in injury-causing crash rates over 7.14M miles to a fleet past 170M miles by late 2025.
First-order effects
- Waymo exits 2016 with both the largest disclosed autonomous mileage base in California and a fourfold improvement in disengagement rate, putting pressure on thin-footprint testers like Tesla, whose 550-mile, four-car program looks experimental by comparison.
- The California DMV gains a working benchmark: with two years of comparable filings, regulators and press can now rank companies on a single normalized number rather than self-reported milestones.
Second-order effects
- Rivals are pushed to scale California testing specifically to stay on the leaderboard — GM's 131K+ miles in the following year's report shows the disclosure regime pulling competitors into the same state rather than private test grounds.
- As rates fall, raw disengagement counts stop flattering anyone; the meaningful comparison migrates toward total miles and, eventually, crash outcomes, forcing every player to publish safety statistics they might otherwise keep internal.
Third-order effects
- If the pattern holds, the disengagement metric retires itself: once fleets run without safety drivers, the industry's credibility shifts from 'how often humans intervene' to independently verifiable injury and crash rates — the framing of Waymo's later safety claims.
- California's role also inverts over time: as operators move into commercial service, in-state test miles shrink (2024 saw a 50% drop), leaving the DMV dataset as a historical record rather than the live measure of autonomy progress.
The trend: Autonomy progress is migrating from state-mandated disengagement disclosures toward operator-published safety outcomes measured over commercial-scale mileage.