Tesla's "10x Safer" Self-Driving Claim Is One Its Own Trainers Don't Believe

A Reuters investigation finds Tesla's headline safety math is inflated — and the very workers who train its "Full Self-Driving" say they wouldn't ride in it. That cuts straight at the robotaxi promise underpinning a $1.6 trillion valuation.

What holds up Tesla's roughly $1.6 trillion market value isn't selling cars — it's a promise: robotaxis. And a late-May Reuters investigation has pried open the most fragile part of that promise's foundation.

Elon Musk and other executives have repeatedly claimed that Tesla's "Full Self-Driving" (FSD) is up to ten times safer than a human driver. The problem is how that number is built. Reuters asked eleven traffic-safety researchers to review Tesla's methodology; ten concluded it amounted to marketing rather than a serious safety study. The reason isn't subtle. Tesla compares the crash rate of FSD-piloted cars that triggered an airbag deployment — that is, serious crashes — against a federal average that folds in far less severe accidents, down to minor fender-benders. It also stacks its relatively new fleet against the average U.S. vehicle, which is considerably older. Normalize those mismatches and the apparent safety edge shrinks dramatically.

More damning is what Tesla's own people think. The investigation drew on interviews with nine former Tesla "data labelers" — the workers who sit in a Utah office reviewing footage from the cars' cameras frame by frame, annotating good and bad driving and escalating problems to engineers. What do they see? Cars striking cats, dogs and deer; failing to brake before impact; speeding routinely; and the occasional near-miss with children playing in the street. Seven of the nine said they would not trust FSD to drive them. A veteran self-driving engineer who spent years reviewing the company's crash data dismissed its safety claims outright and warned others not to take Musk's word on it.

The report surfaces one detail that should unsettle anyone pricing in scale. Ahead of high-profile moments such as robotaxi launches, staff worked long hours hand-training the software on specific routes and specific hazards to make demonstrations look more capable than the system actually is — labor-intensive guardrails that, by the employees' own account, cannot be deployed broadly. In other words, the smoothness under the spotlight is not the same thing as the street.

The contrast with the field's other leader is pointed. Reuters also examined the methodology used by Waymo, Tesla's robotaxi rival, and found it comparatively rigorous — a reminder that the problem isn't that measuring autonomous safety is impossible, only that Tesla's particular framing flatters itself. That distinction matters enormously for a company whose entire thesis rests on removing the human driver: a system that needs route-specific hand-holding to look polished in a demo is, almost by definition, not ready to be turned loose on every street at once. The gap between a curated showcase and a city-wide deployment is exactly where valuations like Tesla's are won or lost.

The regulatory backdrop isn't quiet either. The U.S. National Highway Traffic Safety Administration has four open investigations into FSD and Autopilot, including dozens of cases in which vehicles ran red lights or turned into oncoming traffic. Tesla was also hit last year with a $243 million verdict over a fatal Autopilot crash, which it is appealing. None of this is a verdict on the technology's eventual potential — but it is a long way from the picture Tesla's marketing paints.

For investors, the weight of this story isn't moral; it's narrative. Tesla's valuation today embeds a near-flawlessly executed autonomous future, and much of the recent enthusiasm has rested on the safety statistics this investigation calls into question. The bulls — Cathie Wood, Ron Baron and others — explicitly pin most of the company's worth on robotaxis. Reuters' conclusion runs the other way: on both method and reality, Tesla is nowhere near delivering self-driving safely at scale. One investigation will not reset a stock price overnight, and Tesla has weathered worse headlines before. But it hands the bull case the one thing it most needs and least wants to confront — a concrete, checkable counter-argument, sourced from the very people who train the system. For anyone underwriting the robotaxi thesis, the burden of proof just shifted.

When the spotlight dims and the human guardrails come off, can FSD still carry $1.6 trillion of imagination? That is the question that actually has to be answered.