Xiaomi's Autopilot Warned Too Late, and Three People Died on a Chinese Highway
What happened
In March 2025, a Xiaomi SU7 electric vehicle traveling at 116 kph on the Dezhou-Shangrao Expressway in eastern China struck a cement guardrail while operating in Navigate on Autopilot mode. The collision caused the car to catch fire. The driver and two passengers could not exit. All three died. It was the first major fatality involving Xiaomi's SU7 sedan and the first serious public test of the company's autonomous driving claims.
The standard SU7 did not include LiDAR-based navigation technology. LiDAR adds a layer of spatial sensing that camera-and-radar systems can miss in certain conditions, particularly when a stationary obstacle sits in an unexpected position at highway speed. The autopilot did attempt to respond: it issued a warning and began braking. But the warning came too late for the driver to take control and stop the vehicle before impact. The car hit the barrier before any intervention was possible, regardless of what the driver did after the alert fired.
Xiaomi founder Lei Jun acknowledged publicly that the advanced driver assistance function had been engaged at the time of the crash and said the company would take responsibility, though he did not specify what that meant in legal or financial terms. Xiaomi's stock dropped significantly after the crash became public. The incident arrived at a difficult moment: Xiaomi had been positioning the SU7 as a credible competitor in China's crowded EV market, and its autonomous driving capabilities were central to that pitch.
The deeper problem runs below the specific failure mode. Names like "Navigate on Autopilot" carry strong suggestions about capability. A driver who interprets that name to mean the car can manage highway travel without close supervision is not making an unreasonable inference. If that driver does not find, or does not act on, the technical fine print explaining that the standard hardware lacks the sensors required for every scenario, the gap between what the branding implies and what the system can actually do becomes a design decision with lethal consequences. That gap existed in this car before anyone bought it or drove it.
The crash also exposed a documentation problem that follows this category of technology everywhere it operates. When a vehicle under an assisted-driving system collides with an obstacle, the question of what the system perceived, what it decided, and when it acted determines where responsibility lies. Without a provable record of what a system did in the seconds before impact, liability disputes default to competing claims rather than established facts. Accountability for autonomous driving decisions requires what aviation safety already provides: a verifiable log tied to the system's actual behavior, not reconstructed after the fact from partial and interested reports.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.