NHTSA’s September 21, 2026 openpilot Inquiry: What the Opening Document Says

On September 21, 2026, the National Highway Traffic Safety Administration opened Preliminary Evaluation PE26007 concerning comma hardware and openpilot software. Its opening document described reported crashes involving stopped or slowly moving vehicles and questions the agency intended to investigate.

Research cutoff: October 13, 2026. This article describes the September 21 opening document. Subsequent disposition or status was not verified. It is not a current-status update.

Opening a Preliminary Evaluation is not a final defect determination, completed causal analysis, recall, or civil-liability finding. The useful question here is what the opening actually said—not what later findings might be.

Which products the opening identified

The opening named comma three, comma 3X, and comma four hardware, together with openpilot software. It described aftermarket devices connected to vehicles through harnesses, using cameras and stock vehicle sensors, with features that depend on the software.

That hardware-and-software description matters: “openpilot” alone does not identify the vehicle, installed device, software version, or configuration involved in a particular incident.

Supervised assistance is not driverless operation

NHTSA’s crash-reporting explanation distinguishes Level-2 assistance with speed and steering from Automated Driving Systems spanning Levels 3–5. Level 2 requires a fully engaged human driver; it is not interchangeable with driverless operation.

In its safety instructions, comma describes adaptive cruise control and automated lane centering that require an alert, attentive driver. The developer also says alertness is necessary but not sufficient for safe use. Those are attributed developer instructions—not independent proof of safety, a guarantee that supervision prevents crashes, or an automatic allocation of fault. The changeable instructions reviewed for this article are not verified incident-era documentation.

What the five reported crashes did—and did not—show

The September 21 opening said NHTSA had received notice of five crashes in which vehicles with the devices installed reportedly struck stopped or slowly moving vehicles in the same lane. That is not a finding that openpilot caused five crashes.

NHTSA said it had confirmed systems were “powered on or engaged in multiple” crashes. That phrase does not establish engagement in all five, and it does not provide separate counts for devices merely powered on and systems engaged in assistance. Installation, power, engagement, control authority, and causation are different questions.

The opening also said preliminary data suggest the system may not have adequately detected or responded to in-lane vehicles. Those qualifications are important: the document raised a performance concern for investigation rather than reporting a completed causation finding.

The opening reported two fatal crashes and three fatalities. These were reported outcomes in the incidents under review, not a finding that the system caused those deaths.

There are broader limits to reading crash reports as safety conclusions. NHTSA cautions that reports can be incomplete or unverified and are not normalized for exposure. Raw counts therefore should not be turned into a comparative safety ranking or a crash-risk calculation.

The questions NHTSA intended to examine

The opening document, pages 1–2, described work that included:

• Verifying system status for each reported incident.
• Evaluating performance and safeguards.
• Considering modified openpilot versions, or “forks,” and shared components.

These were intended lines of inquiry at the opening—not findings that the work had been completed. The reference to forks does not establish that every incident involved a modified version, that modifications caused a crash, or that stock and modified versions operate identically.

Keeping the particular hardware, software, vehicle, and configuration distinct helps explain why the opening left incident-specific questions unresolved. A product name alone cannot answer what assistance was active or how it behaved in a given crash.

What this dated opening cannot tell readers

This account does not establish whether PE26007 later closed, advanced, or led to a recall or another action. Follow-up research did not recover a usable later official status record. That access and discovery limit is not evidence that no later action occurred.

For the exact record discussed here, read the September 21, 2026 opening résumé. Its reported incidents and planned inquiry do not resolve civil fault or compensation in an individual case.

Readers interested in automation beyond supervised Level 2 can review the separate evidence questions for driverless ADS vehicles. Those systems raise distinct questions; their operating roles and rules should not be imported wholesale into this openpilot inquiry.

This article provides educational information only, not legal advice.

Source Notes

• NHTSA PE26007 opening résumé, September 21, 2026, pages 1–2: products, reported incidents, tentative concerns, and intended examination. This is the dated record explained here, not a later status determination.
• comma safety instructions: attributed developer guidance; undated and changeable, without a verified incident-era snapshot.
• NHTSA Standing General Order crash-reporting explanation: automation distinctions and reporting limitations. The page obtained during research displayed data through July 15, 2026; it was not an October-current incident dataset.