Tesla Data Points to Driver Override in Fatal Texas Crash

Tesla says vehicle logs from a fatal Katy, Texas crash show the driver pressed the accelerator to 100 percent, manually overriding the car's driver-assist systems.

3 min read
Tesla Data Points to Driver Override in Fatal Texas Crash

AUSTIN, Texas — Tesla has released vehicle data that points away from its driver-assistance software in a fatal crash near Houston, saying logs show the driver manually overrode the system by flooring the accelerator before the vehicle struck a home.

The crash occurred the evening of June 20 in Katy, Texas, where a Model 3 left the road and hit a house, killing 76-year-old Martha Avila. The driver initially told authorities he had been using the car's partially automated driving features. Tesla's review of the car's logs tells a different story, and the company shared its findings as federal regulators opened a routine review. Tesla's safety record with Full Self-Driving has been a central part of its case to regulators worldwide.

What the Logs Show

According to Tesla Head of AI Ashok Elluswamy, the data shows the driver "manually overrode self-driving by pressing the accelerator all the way to 100 percent." The car reached a top speed of about 73 mph, and the accelerator pedal remained fully depressed even after the impact — a pattern consistent with manual input rather than an automated system that is engineered to slow for residential streets.

CEO Elon Musk underscored the point on X, writing that the sequence "makes no sense" if automation were in control, because "FSD drives slowly through neighborhood streets and this was a high speed crash." When a driver presses the accelerator, Tesla's software hands longitudinal control to the person behind the wheel by design, the same way cruise control yields when a driver steps on the gas.

Tesla Data Points to Driver Override in Fatal Texas Crash — additional image

A System Built to Defer to the Driver

The detail matters because it reflects how the technology is supposed to behave. Tesla's driver-assist stack is engineered to keep speeds modest on neighborhood roads and to treat accelerator input as an explicit human command. That architecture is part of why the company has continued to expand the feature set, including the upcoming v14 software improvements rolling out to more of the fleet.

The National Highway Traffic Safety Administration has opened a special crash investigation, a standard step for incidents that draw public attention. A full picture will come once investigators finish reviewing the vehicle's data logs, which Tesla says it is sharing with authorities.

Data Over Narrative

The episode is a reminder of why event data recorders exist. Early accounts of a crash are often incomplete, and detailed telemetry — pedal position, speed, steering, and system state, captured many times per second — offers a far clearer account than first impressions. In this case Tesla says the numbers are unambiguous about who was controlling the car.

Avila's death is a tragedy for her family and community, and the investigation will run its course. But Tesla's willingness to put the logs forward quickly reflects a broader confidence that, as more vehicles record more miles, the data will keep showing its safety systems doing what they were designed to do. TechCrunch's account of the company's response is available here.