How Safe Are Self-Driving Cars?

ADS vs ADAS: Two Very Different Technologies

Understanding self-driving car safety requires distinguishing between two fundamentally different systems:

• ADS (Automated Driving Systems): Fully autonomous vehicles like Waymo robotaxis that drive themselves without human intervention. These operate in geo-fenced areas at lower speeds. 2,977 incidents reported, with 2 fatalities.
• ADAS (Advanced Driver Assistance Systems): Systems like Tesla Autopilot that assist the driver but require human supervision. These operate at highway speeds across the country. 4,105 incidents reported, with 69 fatalities.

Comparing these raw numbers is misleading. ADS vehicles operate millions of miles in controlled urban environments, while ADAS vehicles operate billions of miles in all conditions. The fatality rate per incident for ADAS (1.68%) is higher than ADS (0.07%), but this reflects highway-speed crashes vs low-speed urban incidents — not necessarily system quality. For a detailed head-to-head comparison, see our Tesla vs Waymo analysis.

What the Incident Data Shows

NHTSA's Standing General Order (SGO) requires manufacturers to report crashes involving vehicles with autonomous features. Key findings:

• Severity: 71 fatal incidents, 3,491 injury incidents, and 3,520 with no injuries reported. 50% of all incidents result in no injury.
• Speed: Average pre-crash speed is 36 mph. 2,038 incidents occurred below 25 mph.
• Weather: 3,968 incidents happened in clear weather — most crashes aren't caused by bad conditions.
• Location: Highways (2,008), streets (2,241), and intersections (1,301) are the top locations.

The Missing Metric: Crashes Per Mile

The most important safety metric — crashes per mile driven — is largely unavailable. Manufacturers report total incidents but not total miles driven with autonomous systems engaged. Without this denominator, we can't definitively say whether self-driving cars are safer than human drivers.

Some companies publish voluntary safety reports with mileage data, but these aren't independently verified. Tesla claims Autopilot is safer than average human driving per mile, but researchers have challenged the methodology, noting that Autopilot operates primarily on divided highways — roads that are already safer than average.

Until per-mile data is mandated and independently verified, claims about AV safety relative to human driving should be treated with healthy skepticism — regardless of which direction the claim goes.

The Automation Paradox

One of the biggest safety challenges for Level 2 ADAS systems isn't technical — it's human. The "automation paradox" describes a well-documented phenomenon: the better a system works most of the time, the less prepared the human operator is to intervene when it fails.

When Autopilot handles 99% of driving situations perfectly, the driver becomes complacent and less likely to notice or react quickly to the 1% of situations where the system fails. This complacency risk is a fundamental limitation of Level 2 systems and a key reason why some safety experts advocate skipping Level 2 entirely and going directly to Level 4 (fully autonomous, no driver needed).

What Consumers Should Look For

If you're considering a vehicle with autonomous features, here's what the data suggests:

• Understand the level: Level 2 (ADAS) means YOU are responsible. Level 4 (ADS) means the car handles it in defined areas. There is no consumer Level 5 vehicle.
• Check the safety record: Review manufacturer-specific incident data on AutoPilotWatch to see crash patterns for your vehicle.
• Watch for recalls: Active NHTSA investigations and recalls signal known issues. Check recall status regularly.
• Don't over-trust: Even the best systems fail. The data shows crashes occur in clear weather, on straight roads, at normal speeds. Complacency is the real risk.
• Know the complaints: Check common complaints for your vehicle's system to understand known failure modes.

Common Misconceptions

Public understanding of self-driving car safety is clouded by several common misconceptions:

• "Tesla FSD is fully self-driving" — No. Despite the name, it's a Level 2 system. The driver must remain attentive at all times. See our FSD safety analysis.
• "More incidents = less safe" — Not necessarily. More vehicles on the road means more incidents in absolute terms, even if the per-vehicle rate is lower.
• "No fatalities = safe" — Low fatality counts for some systems reflect small fleets and limited operating areas, not necessarily superior safety.
• "Recalls mean the car is dangerous" — Recalls, especially OTA software updates, show the system is being actively improved. Not all recalls indicate critical safety defects.

The Role of Regulation

NHTSA's investigation process is the primary federal mechanism for ensuring AV safety. Investigations have led to significant recalls and forced manufacturers to improve their systems. At the state level, a patchwork of AV laws creates varying requirements for testing, deployment, and insurance.

Consumer complaints drive regulatory action. Filing a report with NHTSA when you experience unexpected AV behavior is one of the most effective things you can do to improve safety. Every major investigation started with individual complaint patterns.

The Bottom Line

Self-driving car technology is neither the miracle its promoters claim nor the menace its critics fear. The data shows that autonomous systems are involved in crashes — including fatal ones — but that the vast majority of incidents are minor. Whether AVs are net-positive for road safety remains an open question that requires better data than what's currently available.

For now, the safest approach for consumers is to understand exactly what your vehicle's system can and cannot do, stay attentive behind the wheel, and treat autonomous features as tools that assist your driving rather than replace it.