# Uber's test vehicle detected a pedestrian 5.6 seconds before the fatal crash > Detecting something in the road was only one step in the system that was supposed to protect people. Published 2026-11-08 Canonical: https://machines-on-machines.anth.us/posts/uber-tempe-crash-detection-braking/ ::image{src="assets/uber-tempe-crash-detection-braking/uber-tempe-crash-detection-braking.jpg" alt="Investigators examine the Uber test vehicle involved in the Tempe crash" credit="Image: National Transportation Safety Board, U.S. public domain, via Wikimedia Commons."} Federal investigators examine the vehicle after the March 2018 crash in Tempe. [Image source](https://commons.wikimedia.org/wiki/File:UberAutonomousVolvoXC90FatalCrash.jpg). [Licence](https://creativecommons.org/publicdomain/mark/1.0/). In March 2018, an Uber test vehicle struck and killed a pedestrian in Tempe, Arizona. The automated driving system had detected her 5.6 seconds before impact, the National Transportation Safety Board found. It kept tracking her but did not correctly identify her as a pedestrian or predict where she was going. The system's design ruled out using emergency braking to mitigate a collision and relied on the safety operator to intervene. Investigators identified the operator's distraction as the probable cause, while also finding that Uber's risk assessment, supervision and measures to address overreliance on automation were inadequate. The timeline helps explain why a system can register a danger without responding to it in time. Detection had to lead to a prediction and then to an action. The people and procedures surrounding the vehicle were part of that chain. [Read the source](https://www.ntsb.gov/investigations/Pages/HWY18MH010.aspx).