Tesla’s 3.2 Million Cars Under the Microscope: Inside NHTSA’s Largest-Ever FSD Investigation

NHTSA has launched its largest-ever investigation into Tesla's Full Self-Driving software, covering 3.2 million vehicles across all models, after fatal crashes in reduced-visibility conditions raised serious questions about the automaker's vision-only autonomous driving approach.
Tesla’s 3.2 Million Cars Under the Microscope: Inside NHTSA’s Largest-Ever FSD Investigation
Written by Lucas Greene

Tesla just caught a break — and then immediately got hit with something far worse.

Last week, the National Highway Traffic Safety Administration closed a recall investigation into roughly 2.26 million Tesla vehicles over concerns about warning light font sizes. The issue, which involved text that was arguably too small for drivers to read quickly, ended without any mandated recall. A minor victory for the automaker. But the celebration, if there was one, didn’t last long.

NHTSA has now opened what amounts to its most expansive probe yet into Tesla’s Full Self-Driving (FSD) software, covering approximately 3.2 million vehicles across Model S, Model 3, Model X, and Model Y lines from model years 2016 through 2025. The investigation centers on whether FSD’s ability to handle reduced-visibility conditions — fog, sun glare, airborne dust — is fundamentally inadequate, and whether that inadequacy has contributed to fatal crashes.

The scope is staggering. Not just in the number of vehicles affected, but in what it signals about the federal government’s growing impatience with Tesla’s approach to autonomous driving technology.

Four Fatal Crashes, One Common Thread

The probe was triggered by four reported crashes, all involving Teslas operating with FSD or Autopilot engaged, where reduced visibility appears to have been a factor. At least one of those crashes was fatal. According to MSN, NHTSA is specifically examining whether the FSD system can adequately detect and respond to environmental conditions that impair its camera-based sensor array — conditions that any human driver in the same geography would also find challenging, but that a system marketed with the word “self-driving” in its name should arguably handle with particular care.

Tesla’s approach to autonomous driving has long diverged from competitors. Where Waymo and Cruise have relied on lidar, radar, and cameras in combination, Tesla stripped radar from most of its vehicles starting in 2021, betting that a vision-only system powered by neural networks could match or exceed human perception. CEO Elon Musk has repeatedly called lidar a “crutch.” The question NHTSA is now asking, in effect: does a vision-only system become dangerously blind in exactly the conditions where drivers most need backup?

It’s a question the industry has debated for years. Now regulators want a definitive answer.

The four incidents under review reportedly occurred in conditions where camera-based perception would be most stressed. Sun glare washing out a forward-facing camera. Fog diffusing light in ways that confuse object-detection algorithms. Dust reducing contrast between a vehicle or pedestrian and the background. These aren’t edge cases in the traditional software sense — they’re Tuesday afternoon in Phoenix, or any winter morning in San Francisco.

And that’s precisely what makes this investigation different from prior NHTSA actions against Tesla. Previous probes focused on phantom braking, failure to detect stationary emergency vehicles, or the adequacy of driver-monitoring systems. This one strikes at the core sensor architecture Tesla has chosen for its entire fleet.

The agency’s Office of Defects Investigation has classified this as a Preliminary Evaluation, the first formal step in a process that could eventually lead to a recall. But given the 3.2 million vehicles involved, any mandated fix would likely need to come as an over-the-air software update — assuming the problem is solvable in software at all. If NHTSA determines the hardware itself is insufficient, Tesla faces a much more complicated and expensive scenario.

Tesla did not immediately respond to requests for comment. The company dissolved its public relations department in 2020.

The Regulatory Pressure Builds

This investigation doesn’t exist in isolation. It arrives at a moment when Tesla’s relationship with federal regulators is under extraordinary scrutiny, complicated by Elon Musk’s political entanglements and the broader debate over how aggressively the U.S. should regulate autonomous vehicles.

Musk spent much of 2024 and early 2025 closely aligned with the Trump administration, heading the Department of Government Efficiency initiative aimed at slashing federal spending. That proximity raised questions about whether NHTSA — housed within the Department of Transportation — would pull its punches on Tesla enforcement. The opening of this investigation suggests otherwise. Career staff at the agency appear to be proceeding on the merits, regardless of political dynamics.

The timing also coincides with Tesla’s broader strategic pivot. Musk has promised a dedicated robotaxi vehicle, initially slated for production in 2024 but now expected later. The company’s entire bull case on Wall Street increasingly rests on the premise that FSD will eventually achieve full autonomy, transforming Tesla from a car manufacturer into a mobility and artificial intelligence company. A finding by NHTSA that FSD has a systemic problem with common weather conditions would undercut that narrative directly.

Tesla shares have been volatile throughout 2025, buffeted by declining sales in China and Europe, protests at dealerships linked to Musk’s political activities, and growing competition from Chinese EV makers. The stock dropped on news of the investigation, though it recovered partially in subsequent trading sessions. Investors are accustomed to regulatory headlines around Tesla. But 3.2 million vehicles is a number that commands attention even from the most battle-hardened Tesla bulls.

So where does this go from here? NHTSA’s preliminary evaluation could last months. The agency will collect data from Tesla, review crash reports, and potentially conduct its own testing. If it finds sufficient evidence of a defect, the investigation escalates to an Engineering Analysis — a more intensive phase that frequently precedes a recall demand. Tesla can also voluntarily issue a recall at any stage, as it has done numerous times before, often framing software updates as proactive improvements rather than admissions of fault.

The company has issued over a dozen recalls related to Autopilot and FSD in recent years, most delivered via over-the-air updates. In December 2023, Tesla recalled more than two million vehicles to install additional Autopilot safeguards after NHTSA found the driver-monitoring system was insufficient. That recall, too, was delivered as a software patch. But each successive recall chips away at the narrative that FSD is on the verge of solving autonomous driving.

Industry observers note a pattern. Tesla releases FSD updates to its fleet, real-world incidents accumulate, regulators investigate, and Tesla issues a fix. The cycle repeats. What’s changed is the scale and seriousness of the incidents prompting investigation. Fatal crashes in reduced-visibility conditions suggest a failure mode that isn’t easily patched with better driver-attention reminders or tweaked braking parameters. It suggests the system may not see what it needs to see.

For the broader autonomous vehicle industry, the implications extend beyond Tesla. Any regulatory crackdown on vision-only self-driving systems could reshape how companies approach sensor fusion, potentially validating the more expensive multi-sensor approaches used by competitors. It could also slow the timeline for regulatory approval of truly driverless vehicles on public roads — a prospect that would affect Waymo, Zoox, and every startup racing toward autonomous deployment.

But Tesla bears the most immediate risk. 3.2 million vehicles. Four crashes. A federal agency asking hard questions about whether the cars can see through fog.

The answer matters — not just for Tesla’s stock price or Musk’s vision of a robotaxi future, but for the millions of drivers on American roads sharing lanes with vehicles whose software is, by the company’s own admission, still in supervised beta. The “full” in Full Self-Driving has always carried an asterisk. NHTSA is now examining what that asterisk actually means when the weather turns bad and the cameras go partially blind.

No amount of over-the-air updates can fix a problem that doesn’t get acknowledged first. And right now, the federal government is demanding acknowledgment.

Subscribe for Updates

AutoRevolution Newsletter

The AutoRevolution Email Newsletter delivers the latest in automotive technology and innovation. Perfect for auto tech enthusiasts and industry professionals.

By signing up for our newsletter you agree to receive content related to ientry.com / webpronews.com and our affiliate partners. For additional information refer to our terms of service.

Notice an error?

Help us improve our content by reporting any issues you find.

Get the WebProNews newsletter delivered to your inbox

Get the free daily newsletter read by decision makers

Subscribe
Advertise with Us

Ready to get started?

Get our media kit

Advertise with Us