NHTSA Closes Four-Year Investigation into Tesla Phantom Braking
Miles Bennett
NHTSA has formally closed its roughly four-year probe into phantom braking on 2021–2022 Model 3 and Model Y vehicles, finding no pattern of crash risk. This clears both major open safety investigations Tesla faced in the U.S.
What did four years of investigation conclude?
NHTSA found that the unexpected deceleration events on 2021–2022 Model 3 and Model Y vehicles showed no pattern of lane departure or crash risk, and no collision was linked directly to the braking behavior.
This means → regulators determined that phantom braking did occur but never rose to a systemic safety threat — no forced recall is warranted.
The probe began after complaints surged: just 99 in late 2021, jumping to 314 by the time NHTSA opened the case in February 2022. Owners reported their cars braking 10–20 mph without warning in one to three seconds at highway speed.
What actually caused the phantom braking?
NHTSA explicitly ruled out a defect in the automatic emergency braking (AEB) system, calling the two issues separate problems.
The real cause: driver-assistance features — Autopilot, FSD, and Traffic-Aware Cruise Control — misread traffic conditions and triggered unnecessary deceleration.
In plain terms = the brake hardware was fine; the car's software "eyes" misjudged the road and hit the brakes on their own.
Why did complaints spike in 2021–2022 specifically?
The investigation pointed to a pivotal 2021 decision: Tesla switched from a radar-plus-vision fusion perception stack to a vision-only system.
This means → Tesla dropped radar — its "second set of eyes" — and relied solely on cameras. Early in the transition, misread rates climbed and phantom braking clustered.
NHTSA confirmed that both complaint data and field data aligned closely with the timing of that switch.
What happened to complaints after that?
As Tesla iterated its software, complaints fell sharply: 45 for all of 2024, just 3 in the first half of 2026.
In plain terms = the vision-only stack kept updating its algorithms, steadily pushing misread rates down — the problem was largely self-correcting.
One week earlier, NHTSA had separately closed an expanded probe into steering-loss issues affecting roughly 376,000 Model 3 and Model Y units. With both investigations now concluded, Tesla's two major open U.S. safety probes are cleared simultaneously.
What question remains?
Whether the vision-only stack can sustain low misread rates in more complex driving conditions remains the central benchmark for evaluating Tesla's driver-assistance reliability long-term.
This reflects a deeper point: a regulator can close a case, but validating a technology pathway is an ongoing process — it does not end with a single investigation's closure.
Content is for reference only, not financial advice.