Tesla's Own Engineers Called It ‘Radar Saves Us.’ NHTSA Wants to Know Why They Shipped Without It.

NHTSA just sent Tesla a 24-part information demand. Buried inside it is a request for an internal Tesla document titled “Radar Saves Us.” Not a nickname. Not a paraphrase. NHTSA named the document by title, because Tesla itself handed it over during the preliminary investigation and NHTSA wants the full story behind it.^[1]

This engineering analysis covers every Tesla that has ever run Full Self-Driving in the United States. Nine crashes. One dead. One injured. All in the same conditions: sun glare, fog, airborne dust. The cameras went blind, and the car kept driving.

An Engineering Analysis is the last stop before NHTSA either closes a case or orders a recall. There is no step between this and your dealer appointment.

Context matters. In May 2021, Tesla ripped the front-facing radar out of its cars and called the result “Tesla Vision.” Ultrasonic sensors went next, in 2022. Engineers inside the company reportedly argued against the change, warning that cameras alone would fail in rain, glare, and low-light conditions.^[1] Musk overruled them. In private messages later published by Electrek, he conceded the core point: “vision with high-resolution radar would be better than pure vision.” He shipped the cheaper option anyway.

NHTSA isn’t just chasing crash reports. Request No. 12 lists fifteen specific marketing claims and asks Tesla to confirm whether each one “implies a higher level of capability or automation than the system currently has.” The list includes Tesla claiming FSD “can drive in whiteout conditions,” Musk posting that “Tesla FSD can operate in all conditions” over a video that itself admits FSD struggles in fog and direct sunlight, and promotional clips of drivers dancing, brewing espresso, and pushing their seats back while FSD steers.^[1]

Then there’s Request No. 14. In 2024, Tesla changed its “FSD Degraded” warning from red to yellow and killed the audible chime. NHTSA wants every study and every internal justification for softening a warning that tells the driver the system cannot see. That one landed differently when I read it. You dim the alarm specifically when people need it most, and then somebody dies in exactly those conditions.

Tesla has until August 12 to respond. Missing that deadline costs $27,874 per day, capped at roughly $139 million. That cap, by the way, is rounding error on Tesla’s balance sheet. The recall is what would actually cost them.

What you should do right now: If you own a Tesla and use FSD, understand that the system relies entirely on cameras. Cameras can be blinded by glare, fog, dust, and direct sunlight. The car may not alert you loudly enough when this happens. When visibility drops, take control. Don’t trust a yellow warning to tell you what a red one used to. And check whether your vehicle is covered at nhtsa.gov/recalls. If a recall comes from this, you’ll want to be on the notification list.

Limitations: The “Radar Saves Us” document has not been published. Its contents are inferred from NHTSA’s description of why it wants the file. Tesla’s response, due August 12, could reframe or contextualize the document. The nine crashes in the investigation are a small sample; a camera-only system’s broader performance across millions of miles is not captured here. This investigation could still close without action.

Strongest counterargument: Tesla could argue that overall FSD crash rates per mile are lower than human-driven baselines, and that the nine crashes represent an acceptable tail risk. They could also argue the “Radar Saves Us” document was an internal analysis of a deprecated radar configuration, not a verdict on the current vision-only system. If Tesla produces data showing their camera degradation detection improved enough after 2024 to offset the warning downgrade, NHTSA’s case weakens. None of that erases the marketing problem, though.