NHTSA Says Most Waymo School-Bus Violations Came After the Fix
The National Highway Traffic Safety Administration said in a November 6, 2026 filing that a majority of reported Waymo incidents involving illegal passes of school buses occurred after Waymo deployed a software update intended to fix the problem. The agency's Office of Defects Investigation has opened a recall query, a formal review of whether a manufacturer's remedy is adequate, to assess the fix. The NHTSA filing is the agency's most recent public statement on the matter.
Waymo, a unit of Alphabet, said a software update for all affected vehicles was completed on November 17, 2025. That date sits roughly a year after the behavior first drew federal attention. NHTSA's Office of Defects Investigation opened its initial investigation in October 2025, after footage surfaced of a Waymo autonomous vehicle maneuvering around a stopped school bus in Atlanta. In that incident, a robotaxi crossed perpendicularly in front of a bus whose stop arm, the folding red sign that swings out from the side of the bus, was extended. The bus's warning lights were flashing, and children were unloading. The vehicle then turned left around the front of the bus.
The pattern was not confined to Georgia. An earlier incident on September 22, 2025 involved a Waymo vehicle that slowed to 5 mph for a school bus and then proceeded before the bus had retracted its stop arm. Austin Independent School District officials reported at least five incidents of Waymo vehicles illegally passing stopped school buses. In a November 20 letter, the district said five Waymo incidents occurred in November, after Waymo software updates intended to resolve the issue. Earlier reporting from Reuters put the count in Austin at no fewer than 19 illegal passes since the start of the school year. Federal regulators are investigating multiple Texas incidents in which Waymo robotaxis drove around stopped school buses.
A separate incident carries heavier stakes. Waymo said its vehicle quickly detected a child near a school and braked hard. The National Transportation Safety Board, the federal body that investigates transportation accidents, is also investigating the incident, in which a Waymo self-driving vehicle struck a child near a school. The school-bus inquiry and the child-strike investigation involve different federal bodies and different questions, and the two should not be merged into a single finding.
The filing's central point is timing. A software update is only as credible as its field results, and NHTSA's assertion that most reported violations came after the deployment points to a remedy that did not hold in operation. Waymo's position, as stated, is that the update reached every affected vehicle. Those two facts can both be true. The recall query exists to determine which failure mode is driving the incidents: whether the logic was incomplete, whether the perception of stop-arm and flashing-light states was inadequate, or whether the deployed fleet was not uniformly updated in the way the company described.
In my view, the question for operators of driverless fleets is less whether a software fix can be pushed than whether it can be proven in the field against a specific, legally defined hazard. A stopped school bus with its stop arm out is a bounded, well-documented scenario, which makes it a useful test of verification practice. If the recall query produces a clear account of why the remedy fell short, the resulting validation methods could give regulators and operators a repeatable way to confirm fixes for other edge cases before they are relied upon near children.
