Safety is the central promise of autonomous driving: remove the distracted, tired, or impaired human, and you remove most crashes. The reasoning is sound, but measuring whether today's technology actually delivers is harder than it sounds. Here's what we know.
The human-error argument
Government and academic analyses consistently attribute the large majority of crashes to human factors — speeding, distraction, impairment, and simple mistakes. That's the foundational case for automation: a system that never gets tired or distracted could, in theory, prevent many of those crashes. See our ADAS explainer for how today's features already address some of this.
What driver-assist systems show
Studies of advanced driver-assistance systems have found meaningful reductions in certain crash types — particularly the automatic emergency braking systems now standard on many vehicles. The effect is clearest for rear-end and pedestrian-impact scenarios, which is why regulators increasingly treat AEB as essential safety equipment.
What we know about robotaxis
Driverless robotaxi operators publish safety data, but comparing it directly to human driving is methodologically tricky — robotaxis drive in limited, often well-mapped areas and at different hours than typical human drivers. The data is promising but not yet conclusive that fully driverless operation is broadly safer everywhere.
The verdict
Automation has genuine, measurable safety potential — but it's not a solved problem. AEB and related driver-assist features deliver real benefits today, while fully driverless operation is promising but still limited in scope and evidence. Keep your expectations grounded, and read our FSD vs Waymo piece to understand where each technology actually stands.