• NHTSA's Office of Defects Investigation has launched a federal safety probe into Comma.ai's aftermarket driver-assistance system after five reported crashes, including two that killed three people and four that injured up to 11.
  • The agency's initial concern is that the system may not detect or respond to slow or stopped vehicles in the same lane, though it stresses the investigation is not a finding of defect or causation.
  • The probe raises thorny questions about liability for open-source software and modified 'forks,' a key feature of Comma's $999 comma four device and openpilot platform.

Federal Probe Into Comma.ai

The National Highway Traffic Safety Administration has opened a federal safety investigation into Comma.ai, the private automotive-software company founded by George Hotz, following five reported crashes involving its aftermarket driver-assistance equipment. Two of those crashes resulted in three deaths, and four caused injuries to as many as 11 people, according to the agency's Office of Defects Investigation.

At the heart of the probe is a scenario that has long challenged driver-assistance systems: detecting and responding to slow or stopped vehicles ahead in the same lane. NHTSA says the five reported crashes involved failures to detect or react to such vehicles. The agency emphasized that the investigation is not a finding that Comma.ai caused the crashes or that a defect has been established.

The reported cases include a fatal February 2026 crash in Ascension Parish, Louisiana. A 2022 Toyota (7203.T) RAV4, reportedly running FrogPilot—an independently modified, or "forked," version of Comma's openpilot software—struck a stopped emergency-response vehicle, killing two rear-seat passengers. That distinction will likely be central to the investigation: NHTSA is examining Comma-related equipment and software, while modified software can complicate questions of product responsibility.

"What institutional investors like us are really focused on is regulatory stability," said a person familiar with the matter, speaking on condition of anonymity. "But in this case, the regulatory scrutiny is aimed squarely at the product itself."

The Open-Source Liability Question

Comma.ai sells windshield-mounted aftermarket hardware—currently the $999 "comma four"—and develops openpilot, an open-source driver-assistance system that adds adaptive-cruise, lane-centering, and related functions to compatible vehicles. The company says its current product works with more than 325 vehicles and that the driver must remain attentive and ready to retake control.

Comma explicitly says the comma four ships without driving-control software installed and is compatible with "a wide variety of community forks." That ecosystem supports rapid experimentation and broad vehicle compatibility, but it makes software provenance, safety validation, and responsibility after modification more difficult than in a closed, manufacturer-controlled system.

NHTSA will now seek crash logs, version histories, device behavior, driver-monitoring data, installation details, and information about whether official or altered software was used. The agency's findings could set a precedent for how regulators treat aftermarket, software-defined vehicle equipment, particularly when third-party code is involved.

Comma.ai and founder George Hotz had not provided an immediate public response to the report. Hotz is no longer involved in Comma's day-to-day operations, according to the reporting. No verified new executive appointment or corporate restructuring was identified.

Financial and Market Context

As a privately held company, Comma.ai does not publish audited quarterly earnings, revenue, margins, or a current market valuation. Any precise financial-performance figure circulating publicly should therefore be treated cautiously unless it comes directly from the company.

The immediate financial effect is likely concentrated in a niche aftermarket-ADAS market rather than broad public equity markets. Potential near-term costs include legal, engineering, compliance, customer-support, and reputational expenses; sales could also weaken if consumers or insurers perceive the system as unsafe. Conversely, the company's low-price, retrofit model addresses a real market gap: bringing more sophisticated lane-centering and adaptive-cruise functions to cars that did not ship with them.

The news arrives as the automotive sector shifts from conventional ADAS toward increasingly capable—but still imperfect—automated-driving systems. Comma's product is positioned as consumer-accessible assistance, not a self-driving vehicle. That distinction matters: automation can reduce some human-error risks, but drivers may over-trust systems whose marketing or real-world performance seems more capable than their formal operating limits.

This case resembles prior NHTSA scrutiny of Tesla (TSLA)'s Autopilot, particularly crashes involving stationary emergency vehicles. In 2021, NHTSA opened its Tesla probe after identifying crashes in which vehicles using Autopilot struck parked emergency vehicles; in 2023 Tesla recalled more than 2 million U.S. vehicles to strengthen driver-engagement safeguards. NHTSA later said weak driver-engagement controls could enable foreseeable misuse and continued to investigate whether the remedy was adequate.

The parallel does not establish that Comma.ai's system has the same defect or risk profile. It does, however, show NHTSA's established focus: systems that combine permissive operating behavior with insufficient driver monitoring or unclear limitations may receive intensive regulatory scrutiny—especially where stopped emergency vehicles are involved.

Outlook and Implications

Near term, NHTSA is likely to request technical records and determine whether the reports show a recurring safety pattern. Comma may face pressure to clarify official-software support, compatibility with forks, warnings, driver monitoring, and limitations around stationary objects. Users may temporarily reassess installation or usage, while insurers and automakers monitor the investigation.

The result could range from closure without action to voluntary software/hardware changes, a recall, or escalation into a more intensive engineering analysis if the evidence warrants it.

Longer term, the investigation could set an important precedent for aftermarket, software-defined vehicle equipment. Regulators may look more closely at how third-party hardware integrates with OEM safety systems; how modified code is distributed; whether driver-monitoring controls are robust; and what warnings, access restrictions, telemetry, or certification practices are appropriate.

The broader sector's outlook remains mixed: the Department of Transportation believes automation can improve mobility and reduce crashes caused by distraction, fatigue, alcohol, and impairment, but it stresses that the benefits depend on rigorous safety oversight, credible performance standards, and public confidence.

The most responsible current conclusion is therefore narrow: the crash reports raise serious safety questions, especially about detecting stopped vehicles and managing altered software, but the investigation is at an early stage and has not yet determined causation, defect, liability, or remedy.

Correction: An earlier version of this article misstated the number of injuries reported in the crashes. Four crashes reportedly injured up to 11 people, not 10.