Waymo's School Bus Fix Didn't Hold

Federal safety regulators are asking Waymo a pointed question: what did the fix actually fix? According to a November 6 filing, the National Highway Traffic Safety Administration has opened a new inquiry into why Waymo robotaxis continued to illegally pass stopped school buses even after the company shipped a software update meant to end the behavior.

The filing, first reported by Reuters, is a recall query — a distinct step from the defect investigation that preceded it, and one that puts the adequacy of Waymo's remedy directly under review.

A fix deployed last November

Waymo's software update reached all affected vehicles on November 17, 2025, the company said at the time, and it was accompanied by a recall. The update followed numerous reports that its robotaxis were passing school buses that had stopped with their flashing red lights deployed — a maneuver that is illegal for human drivers in most circumstances and, for obvious reasons, dangerous around children.

The recall and the update were presented as the resolution to that problem. Nearly a year later, NHTSA's filing says the behavior has continued.

Most incidents came after the patch

The central finding in the November 6 filing is a temporal one: NHTSA said a majority of the reported incidents occurred after Waymo deployed the software update. That single detail is what turned a closed-out recall into an open question.

The Office of Defects Investigation has opened what NHTSA describes as a “recall query” to “assess the adequacy of the remedy.” In plain terms, the agency is examining whether the software change Waymo made actually corrected the defect it was supposed to correct.

The distinction matters procedurally. A recall query is not a fresh hunt for a defect; it is a review of whether a manufacturer's remedy worked as promised. NHTSA can use the outcome to require further action, but the filing as described does not specify what that action might be.

How the first investigation began

The ODI opened its initial investigation into Waymo in October 2025, prompted by footage of one of the company's autonomous vehicles navigating around a stopped school bus in Atlanta. In that incident, the bus had its stop sign extended, its lights flashing, and was unloading children.

The maneuver captured on video was not a near-miss in the ordinary sense. The Waymo robotaxi crossed perpendicularly in front of the school bus from its right side, then turned left around the front of the bus before continuing down the street.

That sequence — a vehicle routing around an obstacle rather than stopping for it — became the evidentiary seed for the broader inquiry. NHTSA's October 2025 investigation followed from it.

Reports from Austin piled up

Atlanta was not the only location. Numerous other incidents were documented in the weeks and months afterward in Austin, where Waymo also runs a robotaxi service.

Austin Independent School District officials reported at least five such incidents at the time, according to a letter filed with NHTSA. That letter is one of the records anchoring the agency's view of how widespread the behavior was.

The Atlanta and Austin reports together formed the factual basis for the original investigation — and, by extension, for the recall query now examining whether the remedy covered the real-world conditions in which the failures occurred.

What the filing does and doesn't say

The November 6 document is procedural and narrow. It states that a majority of reported incidents occurred after the update. It names the recall query as the mechanism for assessing remedy adequacy. It does not, in the account available, break down the post-update incidents by location, time of day, or the specific configuration of the buses involved.

It also does not describe what the November 2025 software change altered inside Waymo's driving system.

What the filing does establish is a gap between the intended effect of a deployed remedy and the observed record afterward — the precise condition a recall query exists to examine. TechCrunch reported that it had reached out to Waymo and to Austin Independent School District officials for comment and would update its article if they responded.

The remedy under the microscope

Recalls in the automotive world have a well-worn rhythm: defect identified, remedy specified, vehicles serviced, investigation closed. Software-defined vehicles complicate the last step, because the remedy is a code change whose field performance has to be observed rather than measured on a lift.

NHTSA's recall query framework is built for exactly this situation. Rather than reopening the underlying defect question from scratch, the agency is asking whether the remedy performed. The filing's language — assessing “the adequacy of the remedy” — frames the inquiry around outcomes in the field.

Waymo's own account of the timeline is that the update was finished across affected vehicles on November 17, 2025, as previously reported. The recall query turns on what happened next.

School bus stops are a hard test

The maneuver at the center of this story is unusual in the taxonomy of driving behavior. A stopped school bus with flashing red lights and an extended stop sign is not simply an obstacle to be routed around — it is a legal command to halt, with children potentially in the roadway.

The Atlanta footage showed a vehicle that did not halt. It crossed in front of the bus and continued. The Austin reports described a pattern rather than a one-off, with district officials counting at least five incidents at the time of their letter.

For an autonomous system, that pattern is the crux: the same class of situation, recurring, across more than one operating city, after a fix was supposed to have addressed it. NHTSA's filing turns on that recurrence.

Why it matters

The immediate consequence sits with Waymo, but the larger one may land on how self-driving software recalls are handled from here. A recall query that finds a remedy inadequate suggests that shipping a patch and declaring the defect addressed is not, by itself, a sufficient endpoint — and that regulators may increasingly want evidence of field performance before closing a case out.

For operators of autonomous fleets, that implies a longer tail on safety investigations: the period between deploying a fix and being released from scrutiny could stretch, with more reporting obligations in between. For riders, the filing as described does not indicate any service suspension, but it does mean the safety record of the vehicles they ride in remains under active federal review.

For everyone else on the road — especially anyone who drives near a school bus — the open question is narrower and more concrete. A software change was made to stop robotaxis from passing stopped school buses. The agency that oversees vehicle safety says most of the reported incidents came afterward, and it now wants to know whether the change was adequate. How Waymo answers that will shape not just this case, but the standard other automated-driving companies are held to when they promise a fix.