NHTSA Closed the Parking Lot Robot Investigation. It Escalated the Highway One.

On April 3, NHTSA closed its investigation into Tesla's Smart Summon feature. 159 incidents. 97 actual crashes. Zero injuries. Zero deaths. Zero airbag deployments. Case closed, everyone go home. Three weeks earlier, NHTSA upgraded its investigation into Tesla's Full Self-Driving to an Engineering Analysis, the last step before a recall order. That probe covers 3.2 million vehicles, 9 crashes, and one person who is dead.^[1][2]

Two features. Same company. Same vehicles, in many cases. Opposite regulatory outcomes. This is not hypocrisy. This is NHTSA drawing a line that most people haven't noticed: the speed at which your robot operates determines how much forgiveness it gets.

Smart Summon creeps through parking lots at 6 mph. It bumps into gates. It clips parked cars. It fails to see bollards. Out of what Tesla claims are millions of sessions, 97 resulted in any kind of collision. NHTSA's closing document uses the phrase "low incident occurrence and low incident severity" like a doctor telling you the mole is benign. When your robot's worst day is a scratched bumper in a Costco parking lot, regulators shrug.^[1]

FSD operates at 65 mph on roads with oncoming traffic, pedestrians, and sun glare. Nine crashes and one fatality were enough to push NHTSA to Engineering Analysis. The specific failure: FSD doesn't warn drivers when its cameras are blinded by glare, fog, or low sun. The system keeps driving. The human thinks the system can see. Nobody can see.^[2]

What the Highway Robot Has to Beat

FARS data provides the baseline. Tesla Model 3 drivers, operating the same vehicles FSD runs on, kill at a rate of 0.05 deaths per 100 million vehicle miles traveled. Model Y owners manage 0.03. For context, the average American passenger vehicle sits around 1.1.^[3]

Read that again. FSD's human competition isn't the national average. It's Tesla owners, who already crash and die at rates 95% below the fleet average. They tend to be wealthier, drive newer vehicles with advanced safety systems, and crash less often to begin with. FSD isn't trying to replace a drunk in a 2003 Dodge Ram. It's trying to replace a careful driver in a car that already has automatic emergency braking, eight cameras, and a rigid aluminum body.

Total Tesla fatalities in FARS from 2014 through 2023: 278 deaths across all four models. Model S accounts for 100 of those, partly reflecting older model years predating many safety updates. Model 3 contributed 92, Model Y 57, and Model X 29.^[3]

Limitations

FARS does not distinguish between Autopilot, FSD, Smart Summon, and fully manual driving. Every Tesla death in the database is categorized the same way regardless of which features were active. Comparing Smart Summon crash rates (short, low-speed parking lot sessions) with FSD crash rates (extended highway driving) is inherently apples-to-oranges. A 6 mph parking lot bump and a 65 mph highway collision involve different physics by a factor of roughly 100 in kinetic energy. NHTSA's different responses are justified by that gap alone.

What You Can Do

If you own a Tesla with FSD, the NHTSA Engineering Analysis (EA26001) is active. Check nhtsa.gov/vehicle/tesla for updates. Until NHTSA concludes, treat camera occlusion conditions (low sun, fog, heavy rain) as a reason to disengage FSD manually. Tesla's own manual says the same thing. The system will not always tell you when it can't see. If you use Smart Summon, NHTSA considers it low-risk, but stay within line of sight and keep your thumb on the button. Parking lot fender benders won't kill you. They will annoy your insurance company.

Source: NHTSA FARS 2014–2023. Death rates estimated as deaths per 100 million vehicle miles traveled using FARS fatality counts and NHTS-derived fleet VMT estimates. FARS does not tag automation system status at time of crash. See methodology for caveats.