A software update was supposed to stop Waymo’s driverless vehicles from improperly passing school buses. Nearly a year later, federal regulators are questioning whether that fix actually worked.
The National Highway Traffic Safety Administration (NHTSA) has opened a new investigation covering 2,895 Waymo vehicles after identifying continued reports of robotaxis passing stopped school buses with flashing red lights or extended stop arms. The investigation, opened November 6, 2026, follows a December 2025 recall involving more than 3,000 vehicles.
The timing is particularly concerning. Federal investigators say most of the reported incidents occurred after Waymo completed its original software update. With school districts documenting violations and another federal agency examining a separate incident in Texas, the case raises questions about how autonomous vehicles recognize and respond to some of the most vulnerable situations on American roads.
Federal Regulators Open a New Investigation Into 2,895 Waymo Vehicles
The latest investigation marks a significant development in the federal government’s oversight of autonomous vehicles. NHTSA opened recall query RQ26006 on November 6, 2026, focusing on 2,895 vehicles equipped with Waymo’s fifth-generation automated driving system. The agency is examining whether the company’s earlier software remedy adequately corrected a safety problem involving stopped school buses. Unlike a general review of robotaxi performance, this investigation targets a specific requirement: vehicles must remain stopped when school buses display signals indicating that children may be boarding, exiting or crossing nearby. Investigators have documented reports of Waymo vehicles continuing through these situations, sometimes after initially slowing down or stopping.
The new inquiry follows an earlier preliminary evaluation that began October 17, 2025. That investigation examined how Waymo’s technology interpreted school-bus signals and whether it could consistently comply with traffic laws. Although Waymo subsequently issued a voluntary recall, federal officials continued receiving reports of concerning behaviour. NHTSA closed the original evaluation and opened the new recall query after determining that the previous fix required closer examination. Importantly, the investigation does not establish that every vehicle in the 2,895-unit population has committed a violation. Instead, it identifies the systems potentially affected by the issue under review. The central question is whether the software can reliably recognize when remaining stationary is more important than continuing a planned journey.
Why Waymo’s Original Recall Failed to End the Safety Concerns
Waymo’s original recall, filed with NHTSA on December 10, 2025, covered 3,067 fifth-generation automated driving systems. According to the company’s regulatory filing, the problem was linked to software introduced in August 2025 and more broadly deployed the following month. The automated driving system included programming intended to prevent vehicles from unnecessarily obstructing priority vehicles, including buses. However, certain interactions created an unintended conflict. A Waymo vehicle could initially recognize a stopped school bus and yield, only to determine that remaining stationary was obstructing another vehicle. That decision could cause the robotaxi to move before the school bus had switched off its flashing lights or retracted its stop arm.
Waymo engineers identified the behaviour in October 2025, and the company deployed a software correction between November 5 and November 17. Its safety board subsequently approved a voluntary recall after reviewing the issue. At the time of the filing, Waymo reported 12 related traffic citations and said no collisions had resulted from the identified software behaviour. The company considered the update completed, but subsequent reports challenged whether the underlying problem had been fully resolved. NHTSA has now confirmed that a majority of incidents identified during its earlier investigation occurred after the November 17 completion date. That distinction matters because installing a software correction is not the same as demonstrating that the corrected system consistently behaves safely in real-world situations.
Austin School District Documents Repeated Violations Near Students
Austin Independent School District became one of the most vocal critics of Waymo’s school-bus performance after cameras recorded repeated violations during the 2025–2026 school year. In a November 20, 2025, letter to Waymo, the district documented 19 separate incidents involving autonomous vehicles improperly passing stopped school buses. District officials reported that five violations had occurred after Waymo’s November 5 communication indicating that certain software updates were already in place. One particularly troubling incident involved a robotaxi proceeding past a stopped bus moments after a student crossed in front of the vehicle, while the student was reportedly still in the roadway. The district described an average of approximately 1.5 violations per week, raising concerns that the behaviour was recurring rather than isolated.
The school district responded by requesting that Waymo temporarily stop operating during peak student transportation hours, specifically from 5:20 a.m. to 9:30 a.m. and from 3 p.m. to 7 p.m. The request reflected concerns about children who might be entering crosswalks, stepping away from buses or crossing streets where vehicles are legally required to remain stopped. Waymo did not agree to the proposed suspension, according to local reporting. Additional violations continued to be documented, with the district reporting 24 incidents by mid-January 2026. School-bus camera recordings provided an important source of evidence because they captured vehicle behaviour at the moments when warning lights and stop arms were active. The continuing incidents complicated Waymo’s assurances that its software improvements had substantially reduced the problem.
A January Incident Exposes Another Weakness in Autonomous Decision-Making
A January 12, 2026, incident in Austin demonstrated how complicated the interaction between automated driving software and human assistance can become. According to a preliminary report from the National Transportation Safety Board (NTSB), a 2024 Jaguar I-Pace operated by Waymo approached a stopped school bus on East Oltorf Street at approximately 7:55 a.m. The bus was loading students with its red warning lights activated and stop arms extended. Initially, the driverless vehicle behaved appropriately by stopping. It then contacted Waymo’s remote assistance system and asked whether the nearby vehicle was a school bus displaying active signals. An assistance agent located in Novi, Michigan, responded that it was not. The Waymo vehicle subsequently resumed driving and passed the bus while its stop arms remained extended.
The NTSB reported that six vehicles passed the stopped bus during the incident, including the Waymo robotaxi and another vehicle that followed it. No collision occurred, but the sequence highlighted an important challenge for autonomous vehicle developers. A vehicle capable of identifying a potential hazard and stopping can still make an unsafe decision if subsequent information or instructions are incorrect. The presence of remote human assistance does not automatically eliminate that possibility. The NTSB also identified another Austin incident on January 14 involving a school bus operating on a special-needs route. These events remain subject to the safety board’s investigation, and the January 12 sequence should not be treated as proof that the exact software defect identified in the 2025 recall caused every subsequent violation.
Atlanta School-Bus Incidents Reveal That the Problem Extended Beyond Texas
Concerns about Waymo’s school-bus performance were not limited to Austin. In Atlanta, Georgia, school officials documented six incidents involving autonomous vehicles passing stopped buses between May and November 2025. One particularly concerning encounter occurred on September 22, when a Waymo robotaxi approached a stopped school bus from a perpendicular side street. According to NHTSA’s original investigation report, the vehicle initially stopped before maneuvering around the front of the bus and continuing along its opposite side. The bus had flashing red lights, an extended stop arm and a crossing control arm positioned near students who were getting off. No human safety operator was inside the autonomous vehicle during the maneuver, leaving the driving system responsible for navigating the situation.
Waymo subsequently explained that the school bus was partially obstructing the driveway the robotaxi was exiting and that its warning lights and stop sign were not visible from the vehicle’s position. However, the incident illustrated why unusual road configurations can create difficult challenges for automated driving systems. Atlanta Public Schools later confirmed additional violations, while the Atlanta Police Department issued six notices carrying fines of $1,000 each. Local officials also examined how traffic laws could be enforced when a vehicle has no human driver to receive a traditional citation. The Atlanta incidents helped trigger the original federal investigation and demonstrated that problems involving school buses could appear in different cities, even when road layouts and traffic patterns varied substantially.
School-Bus Safety Laws Leave Little Room for Autonomous Driving Errors
School-bus stopping requirements are among the most important traffic protections for children in the United States. NHTSA explains that all 50 states have laws prohibiting motorists from illegally passing stopped school buses displaying flashing red lights and extended stop arms. Although specific requirements vary by state and road configuration, the basic principle is consistent: traffic must stop when required and remain stopped until the appropriate conditions for proceeding have been met. Under Texas Transportation Code Section 545.066, motorists approaching a school bus receiving or discharging students must stop when its required warning signals are operating, subject to specified exceptions. These rules exist because children can move unexpectedly around buses, sometimes entering roadways where approaching drivers have limited visibility.
For autonomous vehicles, complying with these requirements involves more than recognizing the appearance of a yellow school bus. The driving system must interpret warning lights, stop-arm positions, surrounding vehicles, road geometry and the possibility of pedestrians crossing nearby. It must also maintain the correct decision throughout the encounter. A vehicle that stops initially but moves prematurely can create the same danger as a human driver who ignores the warning signals. Waymo’s recall documentation demonstrated how conflicting software priorities contributed to the original defect. The system was intended to avoid obstructing other vehicles, yet that objective sometimes overrode the requirement to remain stationary. The case illustrates why automated driving technology must distinguish between situations where continuing forward is helpful and those where doing so creates an unacceptable safety risk.
Waymo’s Strong Overall Safety Statistics Face a Different Kind of Test
Waymo has repeatedly defended its autonomous driving technology by emphasizing crash statistics that compare favourably with human drivers. The company’s safety dashboard, based on approximately 271.3 million fully autonomous rider-only miles through June 2026, reports substantial reductions in crashes when compared with human-driver benchmarks in its operating areas. Waymo estimates that its vehicles experienced 82% fewer injury-causing crashes and 93% fewer pedestrian crashes involving injuries over the distances examined. The figures reflect the company’s published comparisons and methodology, rather than independent findings that its vehicles perform perfectly in every situation. They nevertheless form an important part of Waymo’s argument that autonomous transportation can improve road safety as the technology becomes more widely available.
Those broader statistics do not eliminate concerns about specific traffic violations. A system can demonstrate lower overall crash rates while still exhibiting behaviour that regulators consider unacceptable around school buses. In fact, the school-bus investigation presents a particularly demanding test because the consequences of a rare mistake could be severe. Waymo has maintained that safety is its highest priority and previously stated that software updates had meaningfully improved performance in these encounters. Federal regulators are now examining whether those improvements were sufficient. For parents and school officials, comparisons with average human drivers may provide useful context, but they cannot replace confidence that a robotaxi will remain stopped when children are entering or leaving a school bus.
The Investigation Could Force Further Software Changes and Intensify Oversight
NHTSA’s latest recall query could have important consequences for Waymo’s operations and its approach to software development. Federal recall investigations can lead to additional corrective actions when regulators determine that an earlier remedy was inadequate. Those actions may include modifying a software fix or expanding the scope of an existing recall. However, the opening of a recall query does not automatically establish a defect or mean that Waymo will be required to remove vehicles from service. Investigators must examine the available evidence, understand the circumstances surrounding reported incidents and determine whether additional safety measures are necessary. For Waymo, the immediate challenge is demonstrating that its automated driving system can consistently comply with school-bus stopping requirements, including in unusual road configurations.
The investigation also comes amid broader scrutiny of Waymo’s performance near schools. In a separate January 23, 2026, incident, a driverless Waymo vehicle struck a nine-year-old pedestrian near an elementary school in Santa Monica, California. The child sustained minor injuries, and federal investigators opened an independent review of that collision. The Santa Monica case is distinct from the school-bus recall investigation, but both highlight the importance of autonomous vehicles responding appropriately around children. As Waymo continues developing and deploying its technology, public confidence will depend not only on favourable crash statistics but also on the company’s ability to identify safety problems, correct them and demonstrate that those corrections work. The newest federal inquiry reinforces a fundamental requirement of autonomous transportation: software improvements must produce dependable safety outcomes on real streets.
