Analysis of all 88 traffic accidents involving autonomous vehicles in California from 2014-2018 shows that autonomous vehicles were at fault in seven accidents
Not Technology—Are the Leading Cause of Self-Driving Car Accidents in California Trevor Mogg / Digital Trends : There's a good reason this self-driving pod has huge ‘virtual eyes’ Tweets: @axios : Since Axios last analyzed the data, there have been 54 new accidents among the 55 companies with self-driving car testing permits in California — most of them continue to be caused by humans. https://www.axios.com/... Justin Green / @jgreendc : People cause most California autonomous vehicle accidents https://www.axios.com/... pic.twitter.com/Onpsn4siUy Scripting Com / @davewiner : 'At the moment, self-driving car tech doesn't seem to be advanced enough to handle all these humans.' http://www.axios.com/...
Context & Ripple Effects
California has been publishing self-driving test incident reports since the DMV first released prototype accident data in 2015, when six early collisions were already attributed mostly to other drivers. The record since then is consistent: Google disclosed 272 disconnections and 13 near-misses over 14 months, and its one possible at-fault event was a 2mph bump against a bus.
Axios's new tally of all 88 accidents from 2014-2018 across the 55 companies holding testing permits — including 54 new crashes since its previous analysis — turns those anecdotes into a dataset: autonomous systems were at fault in just seven cases. That matters because it gives permit holders a quantified rebuttal to safety skepticism, even as later coverage showed firms like Tesla resisting independent verification of their own claims.
First-order effects
- The 55 permitted companies gain a citable statistic for DMV filings and public debate: 81 of 88 recorded accidents were caused by human drivers, not the autonomous systems under test.
Second-order effects
- Rivals' safety marketing now gets benchmarked against this public dataset, pressuring holdouts on disclosure — a gap experts flagged when noting that sparse public data makes it hard to verify Tesla-style safety claims independently.
Third-order effects
- If fault attribution keeps concentrating on human drivers, liability frameworks and insurance pricing shift toward the human side of every collision, strengthening the case for DMV-style mandatory incident reporting as the model other states copy.
The trend: Autonomous-vehicle accountability is moving from anecdotal incident reports to aggregate fault statistics, with California's DMV data becoming the de facto public scoreboard for the industry.