Waymo's autonomous vehicles are involved in significantly fewer crashes than human drivers, according to new research from the Insurance Institute for Highway Safety, though experts caution that standardized data collection remains critical as the technology scales across more cities. The study analyzed crash data from driverless vehicles operating in San Francisco, Phoenix, Los Angeles, and Austin, finding a substantial safety advantage for fully autonomous systems over human-operated cars.

What Does the Safety Data Actually Show?

The Insurance Institute for Highway Safety (IIHS) team determined that self-driving cars were involved in 68% fewer crashes than human-driven cars on a per-mile basis. To reach this conclusion, researchers had to navigate a significant methodological challenge: driverless car companies report incidents using a much lower threshold than human drivers. Autonomous vehicle operators must report any incident resulting in damage, including minor scrapes from turning into a parking lot. Human drivers, by contrast, typically only report crashes to police if they cause injury or more than $1,000 in property damage.

The IIHS team adjusted the driverless car data to match the human driver reporting threshold, creating an apples-to-apples comparison. Waymo vehicles traveled approximately 50 million miles in fully autonomous operation during the study period, compared with about 222 billion miles by human drivers in the same locations over the same timeframe. When researchers applied consistent metrics, Waymo's crash involvement rate was 68% lower than human drivers.

Waymo's own internal reporting aligns with these findings. The company reports that its autonomous vehicles have registered 82% fewer injury-causing crashes and 93% fewer pedestrian crashes with injuries compared to an average human driver over the same distance.

Why Is Waymo Winning the Execution Race?

Waymo's safety advantage stems from fundamental differences in technology and business strategy compared to competitors like Tesla. Waymo treats autonomous vehicles as robotic systems requiring redundant, multi-modal sensory input. A Waymo vehicle uses high-resolution cameras, LiDAR (Light Detection and Ranging), and radar. LiDAR provides an exact, three-dimensional point cloud of the environment regardless of lighting conditions, shadows, or optical illusions.

Tesla, by contrast, relies entirely on a "vision-only" approach using only optical cameras and neural networks. The company stripped radar and ultrasonic sensors from its vehicles to cut costs and streamline manufacturing. While Tesla's vision-only system performs impressively as a driver-assist feature, it struggles with edge cases like glare, heavy rain, and unusual optical anomalies that can confuse the neural network and require human intervention.

This hardware difference has profound regulatory implications. The National Highway Traffic Safety Administration (NHTSA) recently targeted an internal Tesla document titled "Radar Saves Us" as part of an escalating federal probe into Full Self-Driving crashes during low-visibility conditions like thick fog and blinding sun glare. The investigation covers over 3 million vehicles and suggests that even Tesla's own engineering teams recognized critical safety gaps created by removing radar, yet management proceeded with the cost-cutting measure.

Waymo's expensive, bulky, but hyper-reliable sensor suite gave regulators the confidence to remove the safety driver entirely. The company's methodical approach involves mapping cities down to the millimeter and deploying vehicles with redundant sensors, limiting commercial operations to extensively vetted areas. This strategy is expensive and time-consuming, but it actually works safely today.

How Are Waymo and Tesla's Business Models Different?

Beyond hardware, the companies' business models create fundamentally different safety incentives. Tesla sells cars to consumers, meaning liability becomes murky when a privately owned Tesla crashes while using Full Self-Driving. Waymo owns its fleet, assuming 100% of the liability when a Waymo robotaxi operates. This ownership structure forced Waymo to prioritize unassailable safety over rapid deployment.

Tesla's model treats its consumer base as beta testers for autonomous driving technology. You cannot beta-test a commercial robotaxi network on public roads without regulators shutting you down, which is precisely why Tesla's timeline has stalled. Waymo, by contrast, is currently delivering half a million paid, fully autonomous rides per week across more than 10 active U.S. markets. The company has moved from a speculative research and development experiment into a legitimate, revenue-generating logistical utility that citizens rely on for daily commuting.

What Are the Key Differences in Autonomous Driving Approaches?

• Sensor Architecture: Waymo uses cameras, LiDAR, and radar for redundant sensing; Tesla relies solely on cameras and neural networks, creating vulnerability to optical illusions and poor visibility conditions.
• Operational Design Domain: Waymo operates within strictly defined geofenced areas with extensive pre-mapping; Tesla aims for generalized Level 5 autonomy that works anywhere, anytime, which remains technologically unachievable today.
• Liability Structure: Waymo owns its fleet and assumes full liability; Tesla sells vehicles to consumers, creating ambiguous liability when crashes occur and reducing safety pressure on the company.
• Regulatory Approval: Waymo's approach earned regulator confidence to remove safety drivers; Tesla faces federal investigations into crash patterns and internal documents suggesting known safety deficiencies.

What Do Experts Say About Data and Future Safety?

The IIHS study highlights a critical gap: driverless car companies are not uniformly required to report the miles they travel. Waymo is the only company that voluntarily provides this information, making comprehensive safety comparisons difficult.

"As Waymo tackles new weather conditions and traffic conditions in new places, as other companies expand their operations, we need to ensure that this level of safety is maintained. And having more streamlined data is a very important way to do that," said Eric Teoh, lead author of the IIHS study.

Eric Teoh, Insurance Institute for Highway Safety

Teoh noted that while autonomous vehicles are safer than humans on the basis of crash rates, other safety measures and practical deployment challenges require consideration. Jameson Wetmore, an associate professor in the School for the Future of Innovation in Society at Arizona State University, emphasized that no driver is foolproof. "I think anybody who says that a driverless car will not crash is absolutely lying to you or totally misguided," Wetmore stated.

Teoh

Wetmore also highlighted that legal clarity is needed before autonomous vehicle ownership can truly take off. Courts will need to settle questions of responsibility and liability when crashes occur with self-driving cars. Additionally, growth is currently stymied by the lack of an overarching federal organizer for the transition; development is happening piecemeal by private companies.

How to Understand Autonomous Vehicle Safety Levels

• Level 4 Autonomy: The vehicle operates entirely without human intervention within a strictly defined geofenced area or specific environmental conditions, known as an Operational Design Domain. This is where Waymo currently operates and where the proven safety advantages exist.
• Level 5 Autonomy: The vehicle can drive anywhere, anytime, in any weather, exactly as a competent human would, without geographical boundaries. This remains a technological pipe dream in 2026, requiring advances in artificial general intelligence and edge-compute processing that do not yet exist.
• Data Reporting Standards: Autonomous vehicle operators must report all incidents resulting in any damage, while human drivers report only crashes causing injury or over $1,000 in property damage, creating a significant comparison challenge for safety researchers.

The 2026 autonomous vehicle landscape is increasingly defined by execution rather than hype. Waymo's methodical, sensor-rich approach has translated into measurable safety advantages and regulatory approval. Tesla's vision-only strategy, while impressive in controlled environments, has encountered fundamental hardware limitations that federal regulators are now actively investigating. As the technology expands to new cities and weather conditions, standardized data collection and reporting will become essential to maintaining the safety gains that current research documents.