The highway went flat and grey; the lane markers dissolved into glare. You tightened your grip as the car hesitated, then tried to steer itself. I remember the moment I realized this was not a theoretical problem anymore.
On a rain-slick stretch where cameras wash out: 3 Million Teslas Get One Step Closer to a Recall Over Full-Self Driving Feature
The National Highway Traffic Safety Administration’s Office of Defects Investigation announced this week that an inquiry opened in 2024 has moved into an engineering analysis—the penultimate step before a possible recall. That upgrade means regulators will actively test how Tesla’s Full-Self Driving (Supervised) reacts when visibility drops and whether the system warns drivers in time to prevent crashes. If the agency finds a safety defect, the action could touch as many as 3.2 million vehicles.
The complaint targets Teslas running the camera-only Tesla Vision stack introduced in mid-2021. NHTSA says it will recreate low-visibility scenarios, test Tesla’s prior updates and watch for whether cameras detect degraded roadway conditions quickly enough to alert the driver. Nine crashes are tied to the probe so far, including a 2023 fatality that sparked the original field action.
Tesla faces a legal and regulatory squeeze that has been building. A judge this month upheld a $243 million (€224 million) verdict tied to a 2019 Autopilot crash, and a recent Cybertruck suit alleges negligent hiring and design decisions linked to Elon Musk. The California Department of Motor Vehicles has already taken issue with Tesla’s use of the names “Autopilot” and “Full Self-Driving,” calling them misleading; Tesla added “Supervised” to FSD and later sued the DMV.
For drivers, the practical question is simple: which cars are involved? The engineering analysis covers Model 3, Model S, Model X, Model Y and the Cybertruck when they’re running Full-Self Driving (Supervised). I’ve read the NHTSA filing; you should expect targeted tests of those models’ camera systems and of the degradation-detection updates Tesla deployed after the 2023 fatal crash was reported.
The situation reads like a lighthouse without a lamp—navigation tools that promise safety but can fail when conditions get messy. Regulators will be testing whether those lamps can be relit with software updates or whether hardware, labeling and driver alerts need an overhaul.
Will NHTSA force a recall of millions of Teslas?
NHTSA’s engineering analysis is the step that gathers the technical evidence a recall recommendation needs. It doesn’t guarantee a recall, but it raises the probability: if tests show the camera-based system won’t reliably detect degraded road conditions and drivers aren’t being warned, the agency can recommend a recall or other remedies. I’m watching the timeline—these analyses often take months, and Tesla’s public updates will be scrutinized alongside independent testing.
Which Tesla models and systems are under scrutiny?
The focus is squarely on Tesla Vision—the entirely camera-based ADAS rolled out from mid-2021—and vehicles using Full-Self Driving (Supervised). That includes the Model 3, S, X, Y and the Cybertruck when the supervised FSD feature is active. NHTSA will specifically test how the cameras and the degradation detection system handle reduced visibility and whether driver alerts arrive with enough lead time to prevent crashes.
Tesla’s legal troubles and regulatory frictions—court rulings, a high-dollar verdict, consumer watchdog scrutiny and a spat with the California DMV—have built momentum behind the federal probe. The company is defending its naming and development choices while issuing software updates; regulators will judge whether those steps match the risk on the road.
The cameras are the critical sensors today, but if they’re unreliable the system is a brittle scaffold over a busy highway; software patches may not be enough—hardware or labeling changes might be required. So how far will regulators go to fix a problem that could touch millions of cars?