Comma.ai under US federal investigation after two fatal crashes with its hands-free driving system

🕒 Published on Zendoric: September 25, 2026 · 03:17
✨ AI-generated · how it's made
The US National Highway Traffic Safety Administration (NHTSA), through its Office of Defects Investigation (ODI), has opened an investigation into Comma.ai, the startup founded in 2015 by well-known hacker George Hotz.
The U.S. National Highway Traffic Safety Administration (NHTSA), through its Office of Defects Investigation (ODI), has opened an investigation into Comma.ai, the startup founded in 2015 by the well-known hacker George Hotz. The trigger is five reported crashes involving the company's "hands-free" driver-assistance devices, two of which ended in a total of three deaths. In four of the five crashes, as many as 11 people were injured, some seriously.
According to the ODI, the common pattern in the crashes is that Comma.ai's system failed to detect or react to stopped or slow-moving vehicles in the same lane. It is a problem that is not unique to this company: advanced driver-assistance systems (ADAS) have historically struggled with stationary obstacles, and both Tesla and Ford have been the subject of similar investigations in the past.
A relevant detail that adds complexity to the case is that some of the crashes, including at least one of the fatal ones, may have involved "forked" or modified versions of Comma.ai's software — that is, unofficial adaptations created by third parties from the company's open-source code. One of the fatal crashes, which occurred in February 2026 in Ascension Parish (Louisiana), involved a 2022 Toyota RAV4 running FrogPilot, a third-party fork of Comma.ai's "openpilot" software. According to Louisiana State Police, the vehicle struck a Gonzales Police Department unit that was stopped, unoccupied, in the left lane with its emergency lights on while officers investigated another crash. Two passengers riding in the Toyota's rear seats died in the impact. The second fatal crash could not be precisely identified from the report numbers cited by the ODI.
Comma.ai sells devices that can be installed in a wide range of modern cars and that, via its openpilot software, control the vehicle's Adaptive Cruise Control (ACC) and Automated Lane Centering (ALC), with features the company itself compares to Tesla Autopilot and GM Super Cruise. Its motto is "make driving chill," and its website states that openpilot "can automatically steer, accelerate and brake based on other vehicles within its lane." The system also includes a camera-based driver monitor to detect distraction or drowsiness.
Despite these capabilities, Comma.ai explicitly warns users — much as Tesla and GM do with their own systems — that the driver must always be able to retake manual control immediately, either by pressing one of the pedals or hitting the cancel button. The company itself acknowledges on its website that the performance of openpilot's Adaptive Cruise Control can be affected by "vehicles in the same lane that are not moving," and that the system is designed with limits on deceleration and acceleration capability when abrupt braking maneuvers are required.
Neither Comma.ai nor Hotz responded to requests for comment. It is worth noting that Hotz stepped away from the startup's day-to-day management in 2022, so he currently holds no direct operational role at the company.
The case adds to the growing regulatory pressure on semi-autonomous driving systems in the United States, at a time when regulators and manufacturers continue to debate how far these technologies — both those developed in-house by carmakers and those from aftermarket suppliers such as Comma.ai — are capable of safely handling real-world traffic scenarios, especially stopped vehicles or static obstacles. The existence of unofficial software forks, such as FrogPilot, adds a further layer of uncertainty over who bears responsibility when the code actually running on the road differs from the code the company itself develops and validates.
🔗 Related on Zendoric
- Tesla, Microsoft and Anthropic all hit the same wall this week: the AI works, its infrastructure doesn't · 2026-09-08
- The FTC probe lands two days after the safety pact: enforcement, not legislation, is America's AI rulebook · 2026-09-30
- The FTC opens an investigation into the safety risks of OpenAI, Anthropic and other AI companies · 2026-10-02


