Tesla Full Self-Driving Investigation 2026: NHTSA EA26002 Explained

What EA26002 Is and Why It Matters

NHTSA’s Office of Defects Investigation upgraded its Tesla FSD probe from a Preliminary Evaluation (PE24031) to a formal Engineering Analysis on March 18, 2026. The new docket, EA26002, is the second of the agency’s three investigative tiers and the last step before NHTSA can issue an influence letter, demand a recall, or close the case. The Engineering Analysis covers approximately 3,203,754 Tesla vehicles, including 2016–2026 Model S and Model X, 2017–2026 Model 3, 2020–2026 Model Y, and 2023–2026 Cybertruck units running FSD-Supervised software. Background coverage is available from Electrek and Autoweek.

Engineering Analyses do not always produce recalls, but every Tesla recall to date has run through this stage. Gizmodo reported the escalation as the closest a sitting administration has come to a system-wide FSD recall. NHTSA can now demand internal Tesla documents, conduct on-vehicle testing at its New Jersey facility, and issue a formal defect determination. Insurance Journal noted that the investigation directly affects what carriers consider a known risk on these vehicles.

The 9 Crashes That Triggered the Escalation

NHTSA’s public escalation document identifies nine FSD-related crashes where reduced roadway visibility was a confirmed or contributing factor. One of those crashes was a fatal pedestrian strike on November 28, 2023, and at least two of the others involved injuries. Six additional FSD crashes are still under review and may be added to the docket. Each crash listed in EA26002 followed a similar pattern: the vehicle was operating with FSD engaged when sun glare, fog, dust, or another low-visibility condition reduced the camera system’s ability to read the road.

Tesla submitted its Standing General Order crash report on the November 2023 fatality on June 27, 2024, roughly seven months after the crash. The federal Standing General Order requires reports within one day for serious crashes involving Level 2 partial automation systems, and NHTSA has cited the late filing as part of why the agency moved aggressively on this investigation. Tesla’s own analysis acknowledged that a software update rolled out one day later, on June 28, 2024, may have addressed only three of the nine crashes in the docket.

Tesla Vision and the Glare, Fog, and Dust Problem

Tesla’s FSD system runs on Tesla Vision, a camera-only architecture that replaced radar in 2021 and ultrasonic sensors in 2022. Every vehicle in the EA26002 docket relies on cameras alone to perceive the road. NHTSA’s investigation focuses on a specific Tesla Vision shortfall: when sun glare, fog, dust, or other airborne particles reduce camera clarity, the system’s degradation detection sometimes fails to either disengage FSD or alert the driver in time. Industry coverage of this design choice is available from Panter, Panter & Sampedro.

Camera-based perception works well in clear weather. It struggles in low-contrast conditions where human drivers also struggle, but a human driver can squint, slow down, or pull over. FSD does not always make the same call. NHTSA’s data shows the system continued at speed in conditions where a competent human driver would have backed off, leading to crashes that the agency now describes as a probable defect.

The Second FSD Probe: Red Lights and Lane Violations

EA26002 is not the only FSD investigation. In December 2025, NHTSA opened a separate Preliminary Evaluation, PE25012, after collecting 58 consumer complaints describing Tesla FSD running red lights, drifting into oncoming lanes, and crossing into wrong-way traffic. The agency identified 80 separate instances of these behaviors across submitted footage and reports. Tesla’s formal response to PE25012 was due January 19, 2026.

The two investigations cover different defect theories but share the same underlying concern: Tesla FSD is being deployed at scale with known failure modes that the company has not corrected. Where EA26002 focuses on visibility-related failures, PE25012 focuses on traffic-control failures. Together, they represent the most comprehensive federal scrutiny of Tesla’s automation software to date.

The Florida Verdict and Tesla Product Liability

While the federal investigation works through the regulatory process, civil juries are already reaching verdicts. In August 2025, a Miami federal jury returned a verdict of approximately $243 million in a 2019 Tesla Autopilot crash that killed a pedestrian and seriously injured another. The jury awarded $129 million in compensatory damages and $200 million in punitive damages, then assigned 33 percent of the fault to Tesla, with the remainder attributed to the driver. The case is reported in detail by Car and Driver.

The Florida verdict is significant because it broke through a long-standing defense theory that an inattentive driver always cuts off Tesla’s product liability. The jury found Tesla’s system design was a substantial factor in the crash even though the driver was distracted. Plaintiffs’ counsel established that the Autopilot architecture failed to handle a foreseeable crash scenario, which opened the door to comparative fault apportionment against the manufacturer.

Tesla Robotaxi Crash Rate in Austin

Tesla launched a limited robotaxi service in Austin, Texas, in June 2025, running modified Model Y vehicles with FSD-Unsupervised software and onboard safety monitors. Through February 2026, the fleet had logged about 800,000 miles and reported 14 crashes to NHTSA, a rate of roughly one crash every 57,000 miles. Coverage from Bloomberg and analysis by Electrek put that rate at roughly four times the average human-driven crash rate in similar urban conditions.

Robotaxi crashes are reported under NHTSA’s Standing General Order in near real time, which makes them among the most-watched data points in the autonomous vehicle space. The Austin numbers stand in contrast to Waymo’s reported safety record, where the company has logged over 200 million autonomous miles with a far lower per-mile crash rate. The gap suggests the differences in sensor stack, training, and operational design domain are producing materially different real-world outcomes.

Kentucky Product Liability and FSD Crashes

If a Kentuckian is hurt in a Tesla operating on FSD, the case typically involves both the driver and Tesla as a manufacturer. Kentucky’s product liability statute, KRS 411.310, gives manufacturers a presumption against defectiveness when the product was designed within five years before the injury or sold within eight years before the injury. That presumption is rebuttable, and FSD crash cases routinely overcome it through engineering analysis showing the design did not meet the state of the art.

Kentucky follows pure comparative fault for products under KRS 411.182, with the new 50% bar from SB 195 tort reform applying to crashes on or after April 12, 2026. A driver who relied on FSD in conditions where the system was known to fail can still pursue Tesla under product liability, with fault apportioned between the driver, Tesla, and any other contributing party. A breakdown of Kentucky’s product liability framework is available from the Reminger law firm’s state survey.

The statute of limitations for personal injury in Kentucky is one year under KRS 413.140(1)(a), with a separate longer window for product liability claims tied to the date of discovery of the defect. Crash investigation matters more in FSD cases than in conventional crashes because the vehicle’s onboard event data, video, and software logs are critical to proving how the automation behaved at the moment of impact. Tesla preserves that data on its servers, and obtaining it requires fast and specific litigation steps before the company’s retention windows close.

For more on how crash claims are handled in Kentucky, see our Kentucky car accident overview. For additional context on the regulatory environment around tort reform, see our SB 195 breakdown.