Federal safety regulators are investigating Tesla’s Full Self-Driving software after four crashes in conditions where visibility was limited by glare, fog, or dust, including one that killed a pedestrian. The probe, opened by the National Highway Traffic Safety Administration, coincides with Tesla operating driverless taxi rides on rain-soaked Miami streets, raising direct questions about whether the system can handle the weather conditions that triggered the investigation in the first place.
Why NHTSA’s rain and visibility probe puts Miami rides under a spotlight
The tension here is straightforward. NHTSA launched its investigation after documenting four crashes involving Tesla’s Full Self-Driving system in situations where drivers or sensors had reduced sight lines. One of those crashes, detailed in national reporting on the probe, resulted in a pedestrian death. The conditions cited, including glare, fog, and dust, share a common thread with heavy rain: all of them degrade the camera-based perception system that Tesla relies on instead of lidar or radar backup.
Miami’s subtropical climate delivers afternoon downpours for roughly half the year. Operating unsupervised robotaxis in that environment while federal investigators are specifically examining how Full Self-Driving handles poor visibility creates an obvious friction point. If the investigation determines that the software cannot reliably detect obstacles when sensors are compromised, any mandated fix would apply to every Tesla running FSD, including vehicles carrying paying passengers without a human driver ready to intervene.
The hypothesis that a visibility-focused investigation will produce a measurable drop in FSD interventions during rain within six months of a mandated software change depends on two things: whether NHTSA issues a binding corrective action, and whether Tesla’s over-the-air update architecture can deliver targeted improvements to wet-weather perception. Both remain uncertain. NHTSA investigations can take months or years to conclude, and the agency has historically preferred recalls or consent orders over prescriptive software mandates. A quick resolution is not guaranteed, and any timetable for measurable safety gains in rain remains speculative.
At the same time, Tesla has powerful incentives to keep its Miami service running. The company has framed robotaxis as a cornerstone of its future business model, and demonstrating driverless operation in a complex, weather-prone city is central to that narrative. Pulling back service during storms could undermine the claim that FSD is approaching generalized autonomy, yet continuing to operate in heavy rain risks adding new incidents to the very data set NHTSA is scrutinizing.
Four crashes, one death, and the Standing General Order data trail
NHTSA tracks automated-driving incidents through its Standing General Order, a crash-reporting requirement that compels manufacturers to disclose collisions involving vehicles equipped with automated systems. That reporting pipeline is how the agency identified the pattern of four FSD-related crashes under limited visibility. The Standing General Order does not require companies to share raw sensor logs publicly, but it does give NHTSA enough detail to flag clusters and open formal investigations when similar failure modes recur.
The four crashes share a specific characteristic: each occurred when environmental factors reduced what the system could see. Glare from low sun angles, fog banks, and airborne dust all limit the effective range of cameras. Tesla’s decision to remove ultrasonic sensors and radar from newer vehicles means the system depends almost entirely on camera vision processed through neural networks. When those cameras lose contrast or clarity, the system’s ability to identify pedestrians, stopped vehicles, or lane markings can degrade rapidly.
The pedestrian fatality in the crash set carries particular weight. Pedestrian detection is one of the core safety benchmarks that NHTSA evaluates for any automated driving system. A fatal failure in reduced visibility suggests the software may not have adequate fallback behavior when its primary sensors are compromised. That single death converts the investigation from a routine performance review into a potential enforcement action with real consequences for Tesla’s robotaxi ambitions and its claims that FSD can reduce traffic deaths overall.
The Standing General Order data also reveals a structural gap. Manufacturers report crashes after they happen, but the order does not require real-time disclosure of near-misses or disengagements. That means the four documented crashes may represent only the most severe outcomes in a larger pool of degraded-performance events that never resulted in a collision. Without access to Tesla’s internal disengagement logs, NHTSA is working from an incomplete picture of how often FSD struggles in bad weather or how frequently human drivers have to intervene to prevent a crash.
For Miami, that gap is especially relevant. If Tesla’s driverless taxis are experiencing frequent perception dropouts in heavy rain but managing to avoid crashes through conservative fallback behavior-such as pulling over or slowing to a crawl-those incidents may never appear in public NHTSA data. Riders, regulators, and other road users would have no way to gauge how close the system is operating to its safety margins when the skies open up.
Open questions for Miami riders and city regulators weighing robotaxi permits
Several critical pieces of information are missing from the public record. No primary NHTSA crash-report excerpts or raw Standing General Order data have been released that specifically cover Tesla robotaxi operations in Miami rain. The agency has not clarified whether the current investigation includes miles driven by unsupervised commercial taxi vehicles or is limited to consumer FSD activations. That distinction matters because robotaxi passengers, unlike owners behind the wheel, have no ability to take manual control if the system falters.
Miami-Dade regulators have not issued public statements about whether they are coordinating with NHTSA on the probe or imposing weather-related operating restrictions on Tesla’s driverless fleet. Cities that have permitted robotaxi services from competitors like Waymo have sometimes imposed geofencing or weather-based operational limits, such as prohibiting service during heavy rain or dense fog. Whether Miami applies similar conditions to Tesla’s service is an open question with direct safety implications for riders, cyclists, and pedestrians who share the roads with these vehicles.
Tesla itself has not released visibility or weather logs tied to the four crashes cited in the investigation. Without that data, independent researchers cannot assess whether the incidents reflect a systemic software weakness or isolated edge cases. The company’s public communications about FSD have consistently emphasized improving safety metrics over time through software updates, but the specific failure mode under investigation-reduced visibility-is a physics problem as much as a software problem. Cameras have hard optical limits in fog, rain, and glare that no amount of neural-network training can fully overcome, especially when water droplets or dirt physically obscure the lenses.
The recency of the available crash data further complicates the picture. Because Standing General Order reports lag behind real-world deployments, there is no comprehensive, public accounting of how Tesla’s latest software builds perform in Miami’s rainy season. That leaves city officials weighing permits and operating rules with a narrow evidentiary base: a small number of visibility-related crashes, one of them fatal, plus Tesla’s own high-level assurances that each new software release is safer than the last.
For riders, the unanswered questions boil down to informed consent. Passengers booking a robotaxi trip on a sunny afternoon may not realize that the same system is under federal investigation for performance in low-visibility conditions that can arrive with little warning. Clear, plain-language disclosures about the limits of FSD in rain, fog, and glare would give riders a chance to opt out or choose alternative transportation when storms roll in. Absent that transparency, they are effectively test subjects in a live experiment whose parameters they do not fully understand.
Ultimately, the outcome of NHTSA’s visibility probe will shape not only Tesla’s Miami operations but the broader debate over how, and where, automated vehicles should be allowed to drive themselves. If regulators conclude that camera-only systems cannot yet handle the full range of real-world weather, they may push for stricter operating envelopes, additional sensor redundancy, or more robust human oversight. Until those decisions are made, Miami’s rain-slicked streets will remain a proving ground where the limits of Full Self-Driving-and the policies meant to govern it-are still being tested in real time.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
