Waymo's School Bus Fix Faces a Federal Test After Robotaxis Kept Passing

Federal safety regulators opened a recall query on November 6 to see whether Waymo's school bus fix worked. The query covers 2,895 robotaxis. Regulators found that most reports of Waymo cars passing stopped school buses came after the November 2025 software remedy. What they find will shape how software recalls are judged.

• NHTSA opened recall query RQ26006 on November 6, 2026, covering 2,895 Waymo vehicles.

• Waymo completed its school bus software remedy across the fleet on November 17, 2025.

• Waymo filed the related voluntary recall, numbered 25E084, on December 10, 2025.

• Austin ISD counted 19 illegal school bus passes by Waymo vehicles by November 20, 2025.

• The NTSB opened its own investigation of Waymo and school buses in January 2026.

What federal regulators want Waymo to prove

The National Highway Traffic Safety Administration wants proof that Waymo's school bus recall worked. To that end, its Office of Defects Investigation put in a recall query, RQ26006, on November 6 for the 2,895 vehicles with Waymo's fifth gen automated driving system.

In a brief filing, the office notes a large number of reported incidents of Waymo vehicles passing stopped school buses. It says most of the incidents found in an earlier probe took place after Waymo's remedy date of November 17, 2025. A recall query does not amount to a new defect being found; it is the means by which the agency can verify whether a remedy is adequate. The office intends to get to the bottom of the reports that have been coming in since the fix.

How the school bus problem started

It all started in the fall of 2025 with a video out of Atlanta. In it, a school bus is at a standstill with its stop arm extended and red lights flashing as the children disembark. A Waymo crosses in front of the bus from its right side, makes a left turn around its front and is off.

Federal regulators put in motion a preliminary review in October of that year. Then the reports came in from Austin, another of Waymo's robotaxi markets. The Austin Independent School District has on record 19 such cases for the 2025 to 2026 school year. At least five of them occurred after Waymo had informed the district of a software update.

On November 20, 2025, the district's general counsel wrote to Waymo requesting an end to operations during school bus hours. The letter, posted on the NHTSA website, stated the district's view plainly:

Waymo completed the software fix on its entire fleet by November 17, 2025 and put in a voluntary recall, 25E084, on December 10. By January 2026 the National Transportation Safety Board had launched an inquiry into what happened in Austin.

Why is a stopped school bus hard for a robotaxi to read?

A stopped school bus is among the most demanding, rule bound situations on a public road. The driver has to take in the stop arm and the flashing red lights as well as the lay of the land before deciding if the law requires a halt. The answer may turn on the direction of travel, any turns or divided lanes.

Then there are the signals, which can be missed with ease. A stop arm is a small thing and the bus can hide it, and children may step out from places no sensor can see yet. A person behind the wheel is taught to be wary of any yellow bus at rest. With software, someone has to specify when to wait and for how long.

Waymo maintains that its record is solid. According to the company's safety page, in the cities where it operates it has 93 percent fewer crashes involving pedestrian injury than human drivers. In December 2025, when the recall was announced, Waymo chief safety officer Mauricio Peña said in a statement:

That statement now sits next to a federal filing that says the behavior did not change enough. Yet both can be true. A fleet can be safer than average and still fail the one test the public cares about most.

A software fix still has to work on the street

Software recalls are quick to put out, but they are not so easy to prove. Waymo's school bus update was rolled out to the entire fleet in a single move, with no need for parts or a trip to the dealer. What remains to be seen is whether the code has changed the vehicles' behavior in every version of the scene.

It is a familiar story in the industry. Sparknify's review of GM's brake software recall last week turned up the same pattern: the system required a second fix because the 2024 remedy had failed to address the condition. When the defect and the cure are both code, regulators are learning to verify the outcome and not simply the delivery.

Then there is the public cost of a recall query for a robotaxi operator. Waymo is making inroads in new cities, where most streets see school buses twice a day. Each fresh report goes on the record as part of a federal file.

Every fleet faces the same perception problem

Fleets run by humans have the same problem with reading the road, and many now use AI vision for help. A delivery van, bus or truck will pass school buses, crosswalks and traffic lights all day long. The driver is liable to miss something when tired or not paying full attention.

Camera based safety systems for commercial vehicles are meant to close that gap. They can spot a traffic light, keep an eye on blind spots and check the driver for signs of fatigue. It is not like a robotaxi, though: the human stays in control and receives a warning rather than having a decision made for them.

MICROIP, an IC design and AI software firm from Hsinchu, launched its AI Vehicle System Business Group at COMPUTEX 2026 for this very market. The company's automotive grade edge AI unit does front traffic light recognition, rear blind spot warning and in cabin fatigue detection, with 360 degree multi channel recognition for fleet management. At CES 2027, from January 6 to 9, MICROIP can be found in the Taiwan Mobility Pavilion at Booth #4577 in LVCC West Hall. Fleet operators are welcome to book a meeting with MICROIP before the show.

What happens if Waymo's fix still is not enough?

In the near term, federal regulators are likely to keep a closer eye on robotaxis and how they handle children, buses and schools. Should the recall query find the fix is not adequate, Waymo may face another recall and a new software update. Regulators might also call for more data on how its cars behave near schools at pick up and drop off.

Such issues are not confined to one firm. As robotaxis make their way into more cities, they meet different bus routes, road layouts and school timetables. Something that appears resolved in testing can break down on a street unfamiliar to the software. School districts, which collect the violation reports, have become an unexpected safety watchdog.

How the record should be judged is a matter of real dispute. Waymo and its supporters cite the sizable cut in injury crashes compared with human drivers. Critics, the Austin district among them, argue that no fleet should ever go past a stopped school bus, average or not. When the safety of children is at stake, there is little room for patience.

The result has implications for fleets that will never be driverless. Insurers, regulators and school districts are coming to ask whether a system works in a given situation rather than in the aggregate. That standard will shape what buyers ask of every camera and every warning system.

For founders and investors, this is a case in point: autonomy is judged scene by scene. Startups with perception, validation and simulation tools have an obvious pitch, which is to help a fleet show it behaves correctly around crossings, schools and buses before a regulator asks. A company that can put forward such evidence quickly may win contracts that price alone cannot.

Five things will show which way this goes.

1. NHTSA's findings in RQ26006 and whether a new Waymo recall follows in 2027.

2. The NTSB's final report on the Austin school bus incidents, expected by early 2028.

3. School district violation counts in Austin and Atlanta during the spring 2027 semester.

4. Any state or city move to limit robotaxi hours near schools.

5. Whether insurers or regulators begin asking fleets for scene specific safety data.

The next school bus a robotaxi meets will be a test of code, and regulators now want proof that the code passes.

Subscribe to Sparknify for tech news and the people behind it.

Questions & Answers

Q1: I run a fleet of delivery vans with human drivers. With regulators holding Waymo to a stricter standard, can AI cameras help my drivers stop for school buses?

They are of use as a warning, if not a decision. Camera systems on commercial vehicles spot traffic lights, watch blind spots and alert the driver to fatigue. That covers many of the things that can go awry at a bus stop. Fleet operators can book a meeting with MICROIP at sparknify.com/microip-en to see its automotive grade edge AI unit at CES 2027, Booth #4577, January 6 to 9.

Q2: Is the trouble Waymo has with school buses a sign that robotaxis need more sensors, or is the driving software at fault in deciding when to stop?

According to the federal filing, the remedy was a software update rather than new hardware. Waymo already uses cameras, radar and lidar, so the open question is how the system reads a stopped bus. The same split shows up in the fight over camera only self driving, where sensors and software judgment are weighed separately.

Q3: As driverless trucks from companies like Kodiak take to public roads, will federal scrutiny of Waymo's school bus recall slow them down?

Driverless trucks mostly run highway routes, but regulators increasingly want to see a fix work in the scene itself. Autonomous freight firms sell uptime and safety by the mile, as Kodiak's 75 truck deal with DTL shows, and their customers will want the evidence to back it up.