Why federal regulators keep investigating Waymo’s robotaxis around stopped school buses

A driverless Waymo robotaxi struck a 9-year-old walking to an elementary school in Santa Monica, California, on Jan. 23, 2026, and the National Highway Traffic Safety Administration opened a federal investigation five days later. It was the agency’s second preliminary evaluation of Waymo’s behavior around children in four months, and it landed alongside a separate National Transportation Safety Board inquiry into Waymo vehicles driving past stopped, flashing-light school buses in Austin, Texas.

Both investigations followed a recall of more than 3,000 Waymo vehicles for a software defect tied to stopped school buses. Regulators are not chasing one crash; they are chasing a failure that keeps resurfacing after being patched, across two cities and two very different kinds of mistake.

NHTSA’s Two Preliminary Evaluations

The first of NHTSA’s two open cases, designated PE25013, began Oct. 17, 2025, after a media report described a Waymo vehicle driving around a stopped school bus with its warning lights active in Atlanta on Sept. 22, 2025. The agency’s own opening notice said the odds of similar unreported incidents were high, given how much driving Waymo’s fleet does. It cited the company’s roughly 2 million miles of autonomous driving a week.

The second case, PE26001, opened Jan. 28, 2026, after the Santa Monica crash. NHTSA’s notice frames the question narrowly: whether the Waymo vehicle “exercised appropriate caution given, among other things, its proximity to the elementary school during drop off hours.” The child had run into the street from behind a double-parked SUV during morning drop-off; Waymo has said its system detected him as he emerged and braked hard, cutting speed from about 17 mph to under 6 mph before impact. He suffered only minor injuries.

Neither filing is a recall order by itself. A preliminary evaluation is the opening step in NHTSA’s defect process, and if agency engineers find a pattern serious enough, a case can escalate into a full engineering analysis — the same track that, months earlier, had already produced a recall over a related school-bus defect.

The Remote Operator Who Answered ‘No’

Austin, Texas produced a different failure mode entirely. On Jan. 12, 2026, in the 1700 block of East Oltorf Street, a Waymo vehicle approached a stopped Austin Independent School District bus with its red lights flashing and stop arm extended. Rather than stopping, the vehicle’s system queried a remote assistance operator, based in Michigan, with a direct question: was this a school bus with active safety signals? The operator answered no. The vehicle proceeded past the bus.

It was not an isolated event. Austin ISD reported additional incidents in the same window, including one on Jan. 14 involving a special-needs bus route, and school-transportation trade press kept tracking each new report as it surfaced. The National Transportation Safety Board opened a formal docket, HWY26FH007, to examine the pattern rather than a single trip. Its public notice is blunt about how far the review still has to go: “All aspects of the January 12 incident, as well as the other similar incidents associated with Austin ISD school buses, remain under investigation.”

A Recall That Didn’t Fix Everything

Weeks before either incident, Waymo had already recalled part of its fleet over exactly this kind of failure. The company’s recall report, filed Dec. 10, 2025, covers 3,067 vehicles running a version of its fifth-generation driving software released between Aug. 20 and Nov. 5, 2025. The defect: vehicles that had correctly stopped for a school bus with its lights flashing would sometimes start moving again before the bus deactivated those lights. Waymo pushed a software fix between Nov. 5 and Nov. 17, 2025. No collisions were tied to the defect, but regulators noted 12 citations issued to Waymo vehicles during the period it was active.

That fix targeted one specific bug in how the software timed its own stop. It did not touch a remote operator misjudging a bus in Austin two months later, and it did not touch a perception system’s split-second read of a child near a Santa Monica crosswalk five weeks after that. Three separate failure points, three separate investigations, one recall in between that solved none of the ones that followed it.

Two Agencies, Two Different Questions

NHTSA and NTSB are not investigating the same thing, and the distinction matters. NHTSA has enforcement authority: it can order recalls, as it already has, and its preliminary evaluations can escalate into engineering analyses that force a manufacturer’s hand. The NTSB has no such power. It investigates and issues findings and recommendations, and a separate docket, HWY26FH008, covers the Santa Monica crash on its own track, focused on the 9-year-old pedestrian rather than the bus-passing pattern in Austin.

Waymo has said publicly that safety is central to its operation and that its vehicles complete hundreds of thousands of autonomous trips a week in dense urban environments, a scale the company points to when arguing that isolated incidents should be read against a much larger base of safe trips. Regulators reading the same data have drawn a different conclusion: four open dockets across two agencies, tied to a single automated driving system, in under six months.

Each docket is narrower than it sounds. PE25013 is about Atlanta. PE26001 is about Santa Monica. HWY26FH007 is about Austin’s buses. HWY26FH008 is about one Santa Monica child. None of the four has produced a public finding yet, and neither agency has said whether conclusions from one case will inform how it treats the others.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview