Cruise pulled from San Francisco. A setback or

Cruise is done. At least for now.

After a pedestrian was struck by a hit-and-run driver and then dragged 20 feet by a Cruise robotaxi, the California DMV pulled the company’s driverless permit. GM, Cruise’s parent company, then paused all driverless operations nationally. The CEO resigned.

The details matter: the pedestrian was initially hit by a human-driven vehicle that fled the scene. The Cruise vehicle then ran over the already-injured pedestrian, who had been knocked into its path. The Cruise vehicle stopped, detected something under the car, and then moved forward to pull over, dragging the person.

Cruise reportedly didn’t immediately share the full video of the incident with regulators. That omission might be worse than the incident itself.

The plane crash question

I keep asking myself: is this like a plane crash?

Plane crashes are horrific. They generate massive attention. But they don’t stop aviation, because the base rate of safety is so high that individual crashes, while tragic, don’t change the statistical argument for flying.

Is the Cruise incident like that? The technology works most of the time. Autonomous vehicles have driven millions of miles with remarkably few incidents. The statistical case for self-driving is strong: fewer accidents per mile than human drivers.

But “most of the time” sits differently when the failure involves dragging a human body 20 feet.

What’s different about cars

Planes crash in a context where the alternative (not flying) is inconvenient but not dangerous. Cars crash in a context where the alternative (human driving) is also dangerous. 40,000 Americans die in traffic accidents every year. Virtually all of those deaths involve human error.

If autonomous vehicles could cut that number by 90%, the calculus is clear. Save 36,000 lives a year. Accept that a few hundred deaths from autonomous vehicle errors will still occur. The math works.

But the math isn’t how humans process risk. A single dramatic autonomous vehicle incident generates more outrage than 100 human-caused fatal accidents, because the autonomous incident feels wrong in a way that human accidents don’t. We’ve accepted human driving deaths as background noise. Autonomous driving deaths feel like a system failure.

NHTSA data supports the statistical argument for autonomous vehicles. Reuters covered the emotional argument against them. Both are valid. They’re just different kinds of valid.

The trust problem

Cruise’s bigger problem isn’t the incident. It’s the cover-up. Or the perceived cover-up. Not sharing the full video with regulators suggests that someone at Cruise prioritized reputation over transparency.

Trust is the currency of self-driving. Without it, the technology doesn’t matter. You can have the safest car in the world, and if people don’t trust it, they won’t get in.

Cruise damaged the trust. Not just in Cruise, but in the entire self-driving industry. Every Waymo, every autonomous truck, every robot taxi project now operates under slightly more suspicion because one company didn’t handle an incident honestly.

Where this leaves us

Waymo continues to operate. Its safety record is strong. Its transparency has been better. But the regulatory environment just got harder for everyone.

The self-driving future isn’t cancelled. It’s delayed. And the delay isn’t caused by the technology failing. It’s caused by a company failing to be transparent about the technology failing.

That distinction matters. The tech works most of the time. The trust works only when the failures are handled honestly. Cruise broke both, and the industry will pay for it.

Related thinking: