Four‑Second Bloodshed: How Autonomous Driving Algorithms Turned a Fatal Accident

On March 29, 2025, a Xiaomi‑branded SU7 standard‑edition vehicle carrying three female candidates for a civil‑service exam was traveling at 116 km/h on the De‑Shang Expressway under NOA (Navigate‑On‑Autopilot) assistance when it encountered a construction‑zone barrier. Within four seconds the system executed a fatal sequence of "warning‑→‑deceleration‑→‑abandon‑control," causing the driver to lose control two seconds before impact and leaving only 1.5 seconds for a human reaction, ultimately colliding at 97 km/h and igniting a fire that trapped the occupants.

Three key technical failures are highlighted:

1. Night‑vision system becomes "blind" – The low‑cost, lidar‑less visual‑only solution suffers a drastic drop in detection accuracy under low‑light conditions. At 116 km/h the 200‑meter detection range provides merely 6.4 seconds of reaction time, far below the required 10‑second safety buffer for construction zones.

2. Human‑machine handover paradox – Research from Tsinghua University’s Institute of Transportation shows that humans need about 1.5 seconds to perceive, decide, and act, yet Xiaomi’s system transfers control only two seconds before a collision, demanding fighter‑pilot‑level reflexes from ordinary drivers.

3. Emergency‑exit design as "digital violence" – A lithium‑iron‑phosphate battery ignites eight seconds after a 97 km/h impact, exceeding national standards that test at 60 km/h. The electronic lock fails when power is cut, and the mechanical release requires a 20 kg pull force, violating ergonomic principles and contrasting sharply with Tesla’s automatic door‑unlocking after a crash.

The incident serves as a stark industry warning that the cult of automation can be lethal. Marketing narratives that label L2 systems as "quasi‑autonomous" and promise "hands‑free" driving create dangerous dependencies, while current crash‑test regulations overlook algorithmic flaws and leave accident data in a black box controlled by manufacturers.

In conclusion, the tragedy—captured by the victim’s father asking whether they bought a "smart guardian" or a "mobile crematorium"—marks a Chernobyl‑like moment for the smart‑car sector, underscoring that when algorithmic exuberance tramples human ethics, technology can become a weapon of death.