Claims vs. the record
Every operator says it is safer than a human driver, and every launch sounds bigger than it is on day one. This page puts what companies say next to what their own filings show. Both sides are their own words or a government document, linked so you can check either one.
The rules we hold ourselves to
Everything here is the company's own words or an official document, nothing in between. A safety claim can be true and an investigation can still be open; a recall is often just a software fix; an announcement can be honest and simply early. We set the two side by side and let you weigh them. We do not add adjectives.
The full enforcement record →Safety claims vs. the safety record
Each operator's headline safety claim, from its own site, next to the investigations and recalls on file for the same system.
Waymo
Its safety hub says Waymo driving causes 94% fewer serious-injury crashes and 82% fewer injury-causing crashes than an average human over the same miles, and cites a Swiss Re study finding 88% fewer property-damage and 92% fewer injury claims across 25 million miles.
NHTSA has two open investigations into the Waymo driving system, one over hitting stationary objects like gates and parked cars, one over failing to stop for stopped school buses, and Waymo has recalled 1,884 vehicles across two campaigns for colliding with poles, gates, and chains.
Tesla
Tesla's Vehicle Safety Report says vehicles crash far less often per mile when its Autopilot driver-assist is engaged than the US average.
That report measures Autopilot. The software its Robotaxi actually runs on, Full Self-Driving, is under an open NHTSA investigation for running red lights and driving against traffic, 58 incidents, 14 crashes, 23 injuries.
Zoox
Zoox says its purpose-built robotaxi was engineered around safety, with more than 100 safety innovations and equal occupant protection at every seat.
NHTSA opened an investigation after two cases where a Zoox test vehicle braked unexpectedly and a following motorcyclist ran into it, and Zoox recalled 270 vehicles after an April 2025 Las Vegas collision.
Announced vs. actual
What gets announced and what is actually running are often different sizes. Each pair is from the company's or the regulator's own document.
Tesla in Las Vegas
Announced: Cleared to run robotaxis in Clark County, with reports of thousands of cars.
Actually: The Nevada permit itself caps Tesla at ten vehicles, on the Las Vegas Strip corridor, on roads posted no higher than 45 mph.
Waymo's four new cities
Announced: Fully driverless rides turned on in San Diego, Las Vegas, Tampa, and Denver in July 2026.
Actually: Each started employees-first, with public access still described as coming soon at the time of the announcement.
Tesla robotaxi in the Bay Area
Announced: Listed as a Tesla Robotaxi metro.
Actually: Tesla's own Q2 coverage table marks the San Francisco Bay Area 'Safety Driver', a human sits up front, so it is not driverless there.
The disclosure watch
When a number flatters, companies quote it. When it stops flattering, it can quietly disappear. What we are watching:
Alphabet / Waymo
Alphabet told investors Waymo was doing about 500,000 paid rides a week at the first quarter of 2026, then did not mention Waymo at all in its second-quarter release, against a stated goal of a million a week by year end.
Tesla
Tesla publishes a robotaxi coverage map but no fleet size and no ride count, and calls robotaxi not material in its filings, so there is no official number to hold it to.