Waymo's robots keep crash rates well below human drivers, new safety figures show

Waymo is tightening its grip on a core message: its self-driving cars prevent far more accidents than human operators do. The autonomous vehicle firm unveiled fresh safety metrics covering 271 million driverless miles—a significant jump from its previous report—as regulatory battles rage over whether cities should allow more robotaxis onto their streets.

The numbers Waymo is citing#

Across five major cities (Atlanta, Austin, Phoenix, Los Angeles, and the San Francisco Bay Area), Waymo's vehicles recorded substantially fewer crashes involving injuries compared to what statistical models predict for human-driven equivalents:

• 82 percent fewer crashes causing injury
• 95 percent fewer crashes resulting in serious injury
• 93 percent fewer pedestrian crashes with injuries
• 86 percent fewer cyclist crashes with injuries
• 82 percent fewer motorcycle crashes with injuries

The company emphasizes that 271 million miles represents the cumulative equivalent of roughly 300 lifetimes of human driving. These vehicles now operate across more than 10 US cities, completing roughly two million autonomous miles weekly while servicing over 500,000 paid rides per week.

Outside validation—with caveats#

Waymo isn't standing alone with these claims. The Insurance Institute for Highway Safety, an independent research organization, conducted its own analysis and found that Waymo's vehicles experienced 68 percent fewer crashes per vehicle mile traveled than human drivers. However, researchers cautioned that current data-collection methods lack the rigor needed for monitoring a broader rollout of autonomous fleets.

Questions about apples-to-apples comparison#

Skepticism lingers among safety researchers. Critics point out several methodological concerns:

• Uniformity gap: Johns Hopkins University experts note that the industry lacks standardized metrics, making cross-company comparisons unreliable.
• Misleading baseline: Comparing premium, sensor-rich robotaxis against the average American vehicle fleet—which includes older cars and distracted drivers—largely validates that advanced sensors and emergency braking work, not necessarily that autonomous software is superior.