The backup driver of a self-driving Uber that killed a pedestrian in Arizona in 2018 pleads guilty to endangerment, in the first automated vehicle fatality case
Context & Ripple Effects
This closes a five-year arc that began when Uber halted all self-driving tests after its test car struck and killed pedestrian Elaine Herzberg in Tempe in March 2018. The NTSB subsequently found the vehicle failed to identify the pedestrian or brake until too late, and later that it wasn't programmed to recognize jaywalkers at all.
Criminal accountability, however, landed almost entirely on the person in the seat: after a county attorney letter confirmed Uber faced no criminal liability, the backup driver was charged with negligent homicide in 2020. Her guilty plea to endangerment converts that charge into the first resolved criminal case over an automated-vehicle fatality.
First-order effects
- The backup driver's plea ends her criminal exposure under a reduced charge, resolving the only prosecution to emerge from the Tempe fatality while Uber itself remains outside criminal culpability.
Second-order effects
- AV developers now have a template showing the human operator absorbs criminal risk while the company does not — which pressures every operator-supervision program to tighten driver training, monitoring, and handoff protocols before regulators or insurers force the issue.
Third-order effects
- If this allocation holds across future cases, legal accountability for autonomous-driving failures will keep defaulting to the human-in-the-loop rather than the software stack, entrenching safety drivers as both a technical fallback and a liability shield — a structure regulators will eventually have to address head-on.
The trend: As automated-vehicle testing resumes and expands, criminal accountability is settling on individual safety operators while the companies deploying the systems escape it, leaving the division of liability between human and machine unresolved.