March 12, 2026
A dashcam video from West Covina, California went viral Sunday showing a Tesla Model 3 on Full Self-Driving mode crash through active railroad crossing barriers. The timing could not have been worse. March 9 was the day NHTSA’s extended deadline for Tesla to hand over crash video, event data recorder files, and CAN bus data arrived for the second time.
The first deadline was January 19. Tesla asked for more time. The second was February 23. Tesla asked again. The third was Sunday. Tesla is already under pressure on multiple regulatory fronts: European automakers just abandoned Tesla’s EU CO2 compliance pool, stripping more than €1 billion in annual credit revenue. What NHTSA got, and whether it was enough, will determine the next phase of the most consequential autonomous driving investigation in the agency’s history.
2.88 Million Vehicles, One Investigation
Investigation PE25012 covers 2.88 million Tesla vehicles across multiple Full Self-Driving software versions. It’s the largest active NHTSA probe by vehicle count, and the data demands reflect the scale.
The agency’s backlog stands at 8,313 records. Tesla has been processing roughly 300 per day, a pace that would take nearly a month to clear even without new incidents being added. The gap between what NHTSA has asked for and what Tesla has delivered is what drove the two deadline extensions. A third extension would raise serious questions about whether the data NHTSA needs actually exists in the form the agency requires.
The penalties for noncompliance aren’t theoretical. NHTSA can levy fines up to $139.4 million. The more likely escalation path is a formal engineering analysis, which is the step immediately before a mandatory recall order. Tesla has never faced a forced recall on FSD software. That could change.
Fourteen Crashes and Counting
The Austin robotaxi operation, which launched in June 2025, has logged 14 incidents that NHTSA is tracking. The agency has flagged 80 traffic violations across those events, up 60% from its initial count. NBC News independently identified more than 40 railroad crossing incidents involving FSD nationally, suggesting the West Covina video isn’t an isolated failure but a pattern the system hasn’t resolved.
The railroad crossing problem is particularly damaging because it involves a scenario where the consequences of failure are catastrophic. A software system that misjudges a stop sign creates a fender bender. A software system that drives through active railroad barriers creates a potential fatality. The distinction matters for regulators weighing whether FSD in its current form should remain on public roads.
The Pricing Problem Nobody’s Talking About
While the safety investigation dominates headlines, the Austin robotaxi’s business model is quietly struggling. Tesla raised its base fare from $1 to $3.25 this month, a 225% increase. The per-mile rate of $1 stayed the same, but the floor price for any ride more than tripled.
This is the fifth pricing restructure in nine months. The original $1 base fare was a loss-leader designed to build ridership data. Each subsequent increase signals that the unit economics aren’t working at scale. A robotaxi service that needs five pricing experiments in its first year hasn’t found product-market fit. It’s found product-market tolerance testing.
Tesla still plans to expand the autonomous ride-hailing service to Phoenix, Miami, and Las Vegas. Each new city means new mapping, new edge cases, and new regulatory relationships. The Austin experience suggests the path from limited deployment to scaled operation is longer and more expensive than Tesla’s timeline implies.
What Comes Next
NHTSA has three options. First, accept whatever data Tesla submitted by the March 9 deadline and continue the preliminary evaluation. This is the least disruptive outcome and the most likely if Tesla made a good-faith submission. Second, escalate to a formal engineering analysis based on the data gaps and incident pattern. This triggers deeper testing and typically precedes a recall. Third, pursue civil penalties for noncompliance if the data submission was materially inadequate.
The investigation’s resolution will set precedent for every company pursuing autonomous driving. Waymo, which reached its 10-city deployment milestone in February, operates under a fundamentally different regulatory posture: limited geography, dedicated hardware, and incremental expansion. Tesla’s approach of shipping FSD to millions of consumer vehicles and iterating through software updates is faster but carries regulatory risk that compounds with every incident.
The railroad crossing video will fade from social media within days. The 8,313-record backlog won’t. NHTSA’s next move determines whether Tesla’s autonomy ambitions get a longer runway or a shorter leash.
