Self-Driving Car Deaths

Understanding Self-Driving Car Fatalities

As of the most recent data, 71 fatalities have been reported in incidents involving vehicles equipped with autonomous or semi-autonomous driving systems. These fatalities include vehicle occupants, pedestrians, cyclists, and occupants of other vehicles.

It's critical to understand context: the presence of an autonomous system does not necessarily mean the system caused the crash. In some cases, the system was engaged; in others, it was available but not active. NHTSA's reporting requirements capture all crashes involving equipped vehicles, regardless of whether the automation was a contributing factor.

ADAS vs ADS: Where Fatalities Occur

Fatal incidents are disproportionately associated with ADAS (Level 2) systems rather than ADS (Level 4+) systems:

• ADAS fatal incidents: 69 (97%) — primarily Tesla Autopilot at highway speeds where crash energy is highest.
• ADS fatal incidents: 2 — far fewer, reflecting lower operating speeds and smaller fleet sizes.

This distribution makes sense when you consider the physics: a crash at 65 mph involves roughly 7 times the kinetic energy of a crash at 25 mph. ADAS vehicles operating on highways face inherently more dangerous crash scenarios than ADS robotaxis in urban settings. For a detailed comparison, see our Tesla vs Waymo analysis.

Common Fatal Crash Scenarios

Analysis of fatal AV incidents reveals several recurring patterns:

• Stationary object strikes: Vehicles with ADAS engaged failing to detect stopped vehicles, emergency vehicles, or road barriers at highway speeds.
• Road departure: Single-vehicle crashes where the vehicle leaves the roadway, often on curves, suggesting the system lost lane tracking.
• Pedestrian strikes: Fatal pedestrian crashes involving both ADAS and ADS vehicles, often in conditions where detection was challenging.
• Intersection collisions: Fatal crashes at intersections where the system failed to properly assess right-of-way or traffic signals.
• High-speed rear-end: Vehicles rear-ending slower or stopped traffic at highway speeds.

Putting the Numbers in Context

Approximately 40,000 people die on U.S. roads annually. The 71AV-related fatalities, while tragic, represent a tiny fraction of total road deaths. However, this comparison is imperfect — autonomous vehicles cover a small fraction of total U.S. miles driven, so the meaningful comparison would be fatalities per mile, which isn't available.

Each fatality in this database represents a real person and a real loss. The purpose of tracking this data is not to minimize these deaths but to understand patterns that can improve safety. Better data leads to better regulation, which leads to safer systems.

For a broader analysis of whether autonomous vehicles are making roads safer, see our guide on self-driving car safety.

The Role of Driver Attention

A recurring theme in ADAS fatal crashes is driver inattention or over-reliance on automation. NHTSA's investigation of Tesla Autopilot (PE22-002) specifically focused on whether the driver monitoring system was adequate to prevent foreseeable misuse. The finding that it was not led to a major recall.

The challenge of keeping drivers attentive while automation handles most of the driving task is known as the "automation paradox" — the better the system works most of the time, the less prepared the driver is to intervene when it fails. This paradox is at the heart of the debate about whether Level 2 systems can ever be truly safe.