EVwire brief: Tesla VP of AI Software Ashok Elluswamy and CEO Elon Musk publicly disputed reports tying a fatal Texas crash to the company's driver-assistance software, saying the car's own data shows the driver overrode the system.
A Tesla Model 3 left a residential road in Katy, Texas on Friday night and tore through the brick wall of a home, killing 76-year-old Martha Avila, who was inside. The driver, 44-year-old Michael Butler, told Harris County deputies the car was on Autopilot at the time.
However, Tesla's vehicle logs show the opposite. In a post on X, Elluswamy confirmed the driver pressed the accelerator to 100% in the residential area, reached 73 mph (117 km/h), and still had the pedal pinned after impact.
“In this case, the driver manually overrode self-driving by pressing the accelerator all the way to 100% of the accel pedal in this residential area. They reached a speed of 73 mph during the crash, and had the accelerator pressed even after the crash.”
Elon Musk also responded to the story and media claims that Autopilot was involved, stating that the framing of the headlines did not make sense.
The pushback landed because of how the story first reached the public. For two days, the crash was covered under headlines that put Tesla's automated driving system front and center, built almost entirely on Butler's account to the police. As the headlines about the incident continued, the NHTSA stated that it had opened a probe into the crash.
The New York Times ran a headline that read "Tesla Driver Using Autopilot Crashes Into Home, Killing a Woman, Officials Say." NBC News went with "1 person killed as Tesla on autopilot crashes through Texas home." Business Insider's initially read "A Tesla driver using auto assistance crashed into a Texas home, killing a woman," before being updated to “Federal open probe into fatal Tesla crash” following Musk and Elluswamy’s comments on X.
The Times attributed the Autopilot line to officials, who were themselves relaying the driver's claim
NBC stated the Autopilot link more flatly, with no hedge at all.
Some outlets did hedge. ABC News framed the car as "allegedly in driver-assist mode," and Fox News highlighted that "driver says autopilot was on." Hedged or not, however, the headline spotlight landed on Autopilot in every case, and it landed there for one reason: a 44-year-old man, injured and at the scene, told deputies his car was driving itself.
No vehicle data was public at that point. Neither the Harris County Sheriff's Office nor the constable's office had confirmed any technical details about the incident. Investigators said they were still working out why the car failed to control its speed, and what role the driver's own control of it had played.
Elluswamy, for his part, highlighted that preemptive references to Tesla’s technology without confirmed data actually make roads less safe.
Context:
FSD Supervised, Tesla's most capable driver-assistance software, drives cautiously through neighborhoods, slowing for intersections, pedestrians and parked cars. It does not pick a quiet residential road and accelerate to 73 mph. Flooring the accelerator overrides it entirely, the same as in any cruise or lane-keeping system, and the human foot takes over. That behavioral mismatch is the point Musk was making.
Autopilot, the system the driver actually named, is a lane-keeping and cruise feature built for highway driving, not for turning maneuvers through a subdivision. Tesla retired the Autopilot name for new U.S. vehicles in January, though millions of existing cars still run the software, which is part of why the "which system" question is murkier than the headlines let on.
For the record, Tesla's own safety data claims FSD Supervised logs far fewer collisions than human drivers, on the company's own counting method. The same was true for Autopilot before it was retired.
Tesla’s safety stats for FSD Supervised were taken from more than 11.3 billion miles of real-world data
Autopilot deja vu
None of this is new, and that is the part the headlines keep missing. The same script ran in 2021, in this same state.
After a Tesla Model S crashed and burned in The Woodlands, officials declared within hours that no one had been in the driver's seat, and the story instantly became a “fatal Autopilot crash” story. The NTSB found the opposite: a driver was in the seat, and Autopilot could not be engaged on that road at all.
Investigators also found the accelerator had been applied up to 98.8% as the car reached 67 mph (108 km/h) on a residential street. Pedal to the floor in a neighborhood, the same signature now described in Katy.
The similarities between the Katy crash and the Woodlands incident are difficult to ignore. When I investigated this story five years ago, I had a feeling the debunk would get nowhere as much coverage as the false reports. I was right.
The regulators have ruled on this category before, too. Twice in the past five years, NHTSA has reviewed sudden-acceleration petitions covering millions of Teslas, and twice found driver pedal misapplication rather than a vehicle defect.
We have seen this script before, more than once. A careless early claim becomes the story, the data lands days later, and the correction never catches up to the headline. The technology did not drive through that wall at 73 mph. A foot pinning the accelerator did.
Source: Ashok Elluswamy and Elon Musk on X
DIG DEEPER into Tesla's FSD and Autopilot coverage at Teslawire, and join 14,000+ EV geeks getting the day's EV and AV news first.
