LIVE TICKER
[Live wire]Connecting to the newsroom…
TechnologyPublished 10 November 2026 • 10 min read • 2,297 words • 11 cited sources

NHTSA questions whether Waymo's school bus recall fix worked

NHTSA has opened recall query RQ26006 into 2,895 Waymo robotaxis, saying most reports of illegal school bus passing came after Waymo's November 2025 software remedy. The query asks whether the fix worked. No new recall has been ordered.

NHTSA questions whether Waymo's school bus recall fix worked
Wired — view at source ↗

Key facts at a glance

Recall query
RQ26006, opened Nov. 6, 2026NHTSA
Vehicles covered
2,895 Waymo vehicles with the fifth-generation automated driving systemNHTSA
Original recall
25E084, filed Dec. 10, 2025, covering 3,067 vehiclesNHTSA
Remedy completed
Nov. 17, 2025NHTSA
Incidents logged
27 in NHTSA's closing summary, most after the remedy dateNHTSA
Austin school district reports
19 illegal passings reported in Texas by Dec. 4, 2025; five in NovemberReuters

What did NHTSA decide about Waymo's school bus recall?

The National Highway Traffic Safety Administration has escalated its review of Waymo's school bus software. On Nov. 6, 2026, the agency's Office of Defects Investigation closed preliminary evaluation PE25013 and opened recall query RQ26006. The query covers 2,895 Waymo vehicles running the company's fifth-generation automated driving system.

Reuters, TechCrunch and Wired reported the action on Nov. 9. The query began 11 months after Waymo filed the recall it now examines. It also came nearly a year after the company said the remedy was complete.

The trigger is a finding in NHTSA's closing resume. Investigators said most reported school bus incidents happened after Nov. 17, 2025, the date Waymo's remedy was finished. The office wrote that this “calls into question the effectiveness of the recall remedy.”

That sentence is not a finding that the software is still defective. A recall query asks whether a manufacturer's fix addressed the safety problem and whether the recall covered the right vehicles. NHTSA says it aims to complete recall queries within eight months of opening them.

The query's population is narrower than the original recall. Waymo's December 2025 filing covered 3,067 vehicles, while NHTSA's query lists 2,895. Neither document explains the 172-vehicle difference. It could stem from fleet changes, software versions or how each population was counted.

The filings do not suspend Waymo's service or order vehicles off the road, and no new recall has been issued. The query instead puts the burden on Waymo. The company must show its software change worked on public streets, not only in a release note.

The case reaches beyond one company. It tests how federal regulators verify over-the-air fixes for driving software that keeps changing. That question will recur as more robotaxis carry passengers on public roads.

What went wrong in Waymo's school bus software?

Waymo's recall report describes a narrow but serious logic error. A Waymo vehicle could stop or slow for a school bus with flashing red lights or an extended stop arm. It could then start moving again before the bus deactivated its signals. That is illegal in every US state, and it puts children at risk near the curb.

The cause was a conflict inside Waymo's own rules. The system includes features meant to avoid blocking “priority vehicles,” such as public buses and school buses. With its Aug. 20, 2025 software release, Waymo changed how the system predicted other road users' behavior. In some encounters, the vehicle decided it was impeding the school bus and moved to clear the way.

Waymo dates the first known instance of the behavior to Sept. 22, 2025, when a vehicle slowed to 5 mph for a school bus and then proceeded before the stop arm retracted. The event was caught by field monitoring and produced no citation. NHTSA's closing resume cites a media report of a Waymo vehicle passing a school bus in Atlanta on the same date.

The company's chronology shows a rapid response. Engineers identified the flaw in October. Waymo's Field Safety Committee then ordered a hotfix on Oct. 30, outside the normal release cycle. The update reached all affected vehicles between Nov. 5 and Nov. 17. Waymo's Safety Board decided on Dec. 2 to conduct a recall, which the company filed on Dec. 10.

The remedy changed the priority logic. Waymo says the new software stops the system from preferring motion over staying put for a school bus with active signals. The recall report says all 3,067 affected vehicles received the update. Waymo owns every one, so no dealer or owner notifications were needed.

A second failure path appeared in January 2026. The National Transportation Safety Board says a Waymo vehicle in Austin stopped first for a school bus on Jan. 12. Three passenger cars then passed the bus in an adjacent lane. The Waymo asked a remote assistant whether the bus had active signals. An agent in Novi, Michigan, answered “No.”

The vehicle resumed and passed the bus with its stop arms still extended. A passenger car behind it did the same. Six vehicles passed the stopped bus in total, and no crash occurred. The NTSB's case, HWY26FH007, remains open while investigators determine probable cause.

Why is a software recall hard to verify?

A conventional recall ends with a physical fix that inspectors can check. A software recall ends with a behavior change that regulators can only sample. NHTSA must judge whether the revised driving policy handles rare encounters correctly. It relies on data the manufacturer collects and supplies.

Waymo's filing shows how unusual the arrangement is. The company owns every affected vehicle, so there are no owners to notify and no dealers to visit. The remedy was deployed remotely before the recall paperwork was filed. Regulators cannot inspect a part because there is no part to inspect.

What NHTSA can examine is exposure and outcomes. That includes how often Waymo vehicles meet stopped school buses and how often they stop. It also tracks how long they wait and how often they proceed early. Neither the closing resume nor Waymo's recall report gives an encounter count or a post-remedy incident rate. Without those denominators, the public cannot calculate how rare or common the failures are.

The recall query process is built for exactly this gap. A recall query examines whether a remedy was adequate and whether a recall's scope was right. NHTSA lists it alongside preliminary evaluations and engineering analyses.

The stakes are financial as well as safety-related. Wired notes that recall queries have previously led to civil penalties, some costing automakers and software developers tens of millions of dollars. NHTSA can also require a broader recall if it concludes that the original remedy did not work.

Waymo's defense rests on the nature of its product. The company argues that its driving system learns and improves continuously. That sits awkwardly with a recall system designed for fixed parts. Waymo told Wired it looks “forward to continuing our engagement on recall processes for AV software that continues to learn and improve.”

That argument has real force, but it also widens the question. If a system changes every few weeks, a remedy approved in December may not describe the software running a year later. Regulators must decide whether to certify a snapshot or monitor a process. They may also demand proof after each meaningful change.

How many school bus incidents are involved?

NHTSA's closing resume records 27 incidents in its failure-report summary. The office says the majority occurred after Nov. 17, 2025, when Waymo reported the remedy complete. The resume does not publish a separate post-remedy total, an incident-by-incident list or a breakdown by city.

Local reporting fills in part of the picture. Reuters reported on Dec. 4, 2025 that Waymo robotaxis had been reported for illegally passing school buses 19 times in Texas. A Nov. 20, 2025 letter from the Austin district, later posted by NHTSA, said five incidents came in November after Waymo's update. Reuters reported on the letter.

The district kept reporting cases into 2026. The NTSB's docket names a Jan. 14 incident involving a 2023 International school bus on a special-needs route, alongside the Jan. 12 case. The Jan. 12 bus was equipped with external cameras, including stop-arm cameras that record vehicles failing to stop.

Camera footage is central to the evidence. NHTSA's preliminary evaluation was prompted by media reports. Waymo's earliest recorded instance of the behavior is the Sept. 22, 2025 event. That matches the date of the Atlanta footage cited by the agency. Texas law treats passing a stopped school bus with active signals as a violation under Transportation Code Section 545.066.

Citations are a narrower measure than incidents. Waymo's recall report says the company knew of 12 citations related to the behavior, all issued between Sept. 25 and Nov. 14, 2025. A citation requires an officer to witness or document the violation. The count almost certainly understates how often the maneuver happened.

The documents also record no physical harm from these events. Waymo's recall report says no collisions occurred, and NHTSA's incident table lists no crashes, fires or injuries. That is relevant context, but it does not settle the safety question. A vehicle passing a stopped bus can injure a child without striking anything first.

The numbers matter because the query hinges on timing. If incidents cluster before the remedy and fade afterward, the fix worked. If they continue at a similar pace, the remedy failed. NHTSA has not published the data that would let outsiders draw that conclusion.

What does Waymo say about the investigation?

Waymo disputes the idea that it has not addressed the problem. Spokesperson Dana Gallagher told Wired that the company has “made significant improvements to vehicle behavior around school buses.” Gallagher said Waymo has not been cited for passing school buses since March 2026, months after the December recall.

The company also points to its broader safety record. Waymo said in September 2026 that its vehicles had driven more than 270 million fully autonomous miles through the end of June. The figure covers Atlanta, Austin, Los Angeles, Phoenix and San Francisco. It reported an 82% reduction in injury-causing crashes compared with human benchmarks.

Waymo's analysis also claims a 93% reduction in injury-causing crashes involving pedestrians. Those figures come from the company's own methodology. It compares its fleet against human benchmarks. They are not an independent audit, and they do not address school bus stop-arm compliance directly.

The company has taken practical steps. Wired reported that Austin school officials helped set up a half-day data collection event after the recall. School buses and stop-arm signals were brought to an empty parking lot so Waymo could train its vehicles on the maneuver. Incidents still followed, according to the district's reports.

Waymo has not released encounter-level data on how its vehicles behave around stopped buses. The gap covers behavior before and after the remedy, and it makes the company's improvement claim hard to assess.

There is a commercial reason for caution on both sides. Waymo's business depends on public trust in its safety claims. A prolonged regulatory fight would undercut that trust. NHTSA, meanwhile, must show that software recalls have consequences. Both have incentives to describe the same data in favorable terms.

What are the limits of the evidence?

The strongest fact in NHTSA's resume is also the bluntest: most incidents came after the fix. That is a timing comparison, not a rate. It does not by itself prove the remedy failed. Reporting may have improved after the recall, and the fleet may have grown.

Reporting bias cuts both ways. A district that installs stop-arm cameras and reviews footage will find more violations, including ones that predate the recall. Publicity after December 2025 may have made it easier for drivers and school staff to report Waymo vehicles. More reports do not always mean more violations.

Citation data is thinner still. Police must see the maneuver or obtain footage quickly enough to identify the vehicle. Waymo's claim that it has received no citations since March 2026 is therefore reassuring but incomplete. It measures enforcement, not compliance.

The software itself remains a black box. NHTSA can demand data, but outside researchers cannot audit Waymo's perception and planning code. The public must rely on summaries from the company and the agency. Neither has released the underlying event logs.

A separate NTSB case adds context without answering the recall question. On Jan. 23, 2026, a Waymo vehicle traveling at 17 mph in Santa Monica struck a student pedestrian. The pedestrian had crossed between vehicles near 24th Street and Pearl Street. The pedestrian reported minor injuries and did not need medical transport.

That crash is not a school bus passing case, and investigators have not linked it to the software defect. The NTSB says probable cause in both investigations is still undetermined. Neither case has published findings, and any recommendations will come with the final reports.

The fair conclusion is that NHTSA has a credible signal and an incomplete case. The agency is right to ask whether the remedy worked. It has not yet shown, in public, that the software still fails at a rate that warrants another recall.

What happens next for Waymo and robotaxi regulation?

Waymo must answer NHTSA's query with data on its software, its fixes and its school bus encounters. NHTSA's eight-month target for recall queries would point to a decision around mid-2027. That is a target, not a deadline.

Several outcomes are possible. NHTSA could accept the remedy and close the query. It could expand the recall to more vehicles or require a different fix. It could also seek civil penalties. Wired notes that past recall queries have produced such penalties. None of those steps has been announced.

The NTSB's work runs on a separate track. Its investigations usually end with safety recommendations rather than enforcement. Recommendations on school bus detection, remote assistance protocols or stop-arm camera data could reach regulators and other robotaxi operators. They would not be binding.

The most important evidence will come from the current school year. If Austin and other districts report fewer incidents through spring 2027, Waymo's case strengthens. If reports continue at the same pace, NHTSA will face a harder question. It will have to decide whether the problem is the software, the training data or the way remote assistants answer questions.

Watch the scope gap between 2,895 and 3,067 vehicles. If NHTSA concludes that the smaller population excludes vehicles that still have the old behavior, the query could widen. Watch, too, whether the agency publishes encounter rates rather than raw incident counts.

Other operators are watching closely. Every robotaxi company that uses over-the-air updates will face the same question. How does a regulator verify a fix for software that never stops changing? Waymo's case will set an early precedent, whichever way it ends.

For parents and school officials, the practical advice is unchanged. Report stop-arm violations with footage and a time stamp to both the company and NHTSA. The Atlanta footage and the Austin district's citations helped open this case. Similar reports will shape how quickly the next one is resolved.

Books behind the context

  • Lawrence D. Burns with Christopher Shulgan

    An insider account of Google's self-driving program that explains the safety-validation culture Waymo inherited and the gap between lab testing and public streets.

  • Hod Lipson and Melba Kurman

    Describes the perception and planning stack behind automated driving, which helps explain why a stop arm and a flashing light can still confuse a trained system.

  • Anthony M. Townsend

    Examines how automated vehicles collide with street rules, public institutions and oversight, the terrain this recall query is now testing.

Sources used for this dispatch

  1. NHTSA Office of Defects Investigation - PE25013 closing resume and RQ26006 opening
  2. NHTSA - Part 573 Safety Recall Report 25E084 (Waymo)
  3. NTSB - HWY26FH007: Automated driving system vehicle passed a school bus loading students
  4. NTSB - HWY26FH008: Automated driving system vehicle collision with student pedestrian
  5. Wired - Waymo Said It Fixed Its School Bus Problem. The US Government Says It Didn't Work
  6. TechCrunch - Waymo robotaxis kept illegally passing school buses after software fix, feds say
  7. Reuters - US opens new probe into 2,895 Waymo vehicles over school bus safety violations
  8. Reuters - US probes reports Waymo self-driving cars illegally passed school buses 19 times in Texas
  9. Reuters - US safety board opens probe into Waymo robotaxis passing stopped school buses
  10. Waymo - From the road: safety data, September 24, 2026
  11. NHTSA - Resources related to investigations and recalls

This article is original analysis and synthesis. Facts, figures and quotations are attributed to the sources above. Images remain the property of their publishers and are credited in the caption.

Questions readers ask

What is NHTSA recall query RQ26006?

RQ26006 is an investigation into whether Waymo's December 2025 software recall actually fixed its school bus passing defect. NHTSA's Office of Defects Investigation opened it on Nov. 6, 2026, covering 2,895 vehicles, and aims to finish recall queries within eight months.

How many Waymo vehicles are covered?

The query covers 2,895 vehicles with Waymo's fifth-generation automated driving system. The original recall, 25E084, covered 3,067 vehicles. Neither document explains the difference between the two figures.

Did Waymo's software fix fail?

NHTSA says most reported school bus incidents happened after Waymo's remedy was completed on Nov. 17, 2025, and that this calls the remedy's effectiveness into question. The agency has not issued a formal finding that the fix failed.

Is Waymo being forced to recall its robotaxis again?

No. RQ26006 is a review of the existing recall, not a new recall order. NHTSA has not stopped Waymo's service. The query could end with the remedy accepted, a wider recall, a new fix or civil penalties.

Where did the school bus incidents happen?

Austin, Texas, is central to the case. Reuters reported 19 illegal passings in Texas by Dec. 4, 2025, and a district letter said five came in November after Waymo's update. NHTSA also cites a Sept. 22, 2025 incident in Atlanta, and the NTSB is examining cases from Jan. 12 and Jan. 14, 2026.

Is Waymo still operating while the query is open?

Nothing in the NHTSA filings suspends Waymo's service. Waymo reported operations in Austin, Atlanta, Phoenix, Los Angeles and San Francisco in its September 2026 safety update, and NHTSA has not ordered a stop to service or issued a new recall.

Topics:WaymorobotaxiNHTSAschool bus safetyautonomous vehiclessoftware recall

More from the research desk