Federal auto safety regulators are pushing Waymo for extensive answers after Austin school officials alerted the company that its driverless taxis were repeatedly zooming past stopped school buses. The examination, which is being overseen by the National Highway Traffic Safety Administration’s Office of Defects Investigation, focuses on how Waymo’s system understands stop-arms and flashing lights — signals that are intended to cause every driver to come to a stop.
What NHTSA Wants to Know About Waymo’s School Bus Cases
Investigators used a formal information request this week to ask Waymo to explain the school bus behavior of its fifth-generation automated driving system, outline any software mitigations, and describe operational safeguards in Austin. Regulators also asked whether Waymo had suspended operations during opening or closing hours at each school at the district’s request, whether recent software upgrades have fixed the hazard, and if the company wishes to issue a safety recall.
ODI’s procedure can proceed from opening inquiries or information collection to an engineering analysis and, as necessary, a recall. For carmakers, a “recall” for automated driving systems usually involves an over-the-air software fix, but the move still must include a public defect filing under federal rules and clear-cut evidence that the problem is isolated.
19 Illegal Passes Reported by Austin Schools
Austin Independent School District informed Waymo that it had recorded 19 alleged instances this school year in which driverless cars went around buses flashing stop-arms and red lights. Several of those reports were received after Waymo began rolling out an update to its software that aimed to help the vehicle perform better around school buses, and so district officials urged the company to stop operations during morning and afternoon pickup times.
According to the Texas Transportation Code 545.066, drivers must stop for school buses with their red lights flashing and may not continue until the bus starts moving or the lights turn off, or a traffic officer decides it is safe.
The fines are hefty because the risk factor — children running out from in front of or behind the bus — is particularly high.
Unless That Didn’t Work: A Previous Incident and a Software Fix
Waymo’s school bus performance was put through the paces after video showed a company vehicle circumnavigating a stopped bus in Atlanta during student unloading. Waymo said the bus had blocked parts of its approach to its driveway and obstructed the stop-arm and flashing lights, causing the system to interpret it more as a roadblock than a required stop. The company has been responding with a software update designed to improve detection and right-of-way assignment, specifically when it comes to school buses.
Waymo emphasized its commitment to safety; the company said that five times fewer injury crashes occurred than with human drivers and 12 times fewer for pedestrians, citing internal data in a statement. The company said it is working with NHTSA and argues that the most recent updates “substantially” improve behavior around school buses, and now exceed typical human compliance in these edge cases.
Reuters was first to report on the federal letter seeking more information from the Alphabet-owned company. The agency’s questions indicate that regulators are considering not just the technical fix, but whether or not to impose operating limits during software learning.
Why School Buses Are an Issue for Robotaxis
Stopped school buses have long been a test case of perception and policy. Stop-arms can be partially obscured, lights may only be visible at an angle or due to glare, and vulnerable road users may be around but not immediately obvious. Being conservative — considering any school-bus-like cues to be the same as a hard stop — reduces false negatives but risks blocking traffic when signals are not active. There is too much of a permissive attitude: the unacceptable can pass when children are crossing.
Developers can combine solutions, such as high-resolution vision and lidar fusion for stop-arm detection, map-level “no-passing” logic at known bus stops, or time-of-day slow zones near schools. Vehicle-to-everything pilots that enable buses to send a message with a “stop” notification directly to vehicles in their immediate vicinity could be transformative, but coverage is limited and standards are yet to solidify. Until then, the standard is still strong edge-case handling and conservative motion planning.
What Happens Next for Waymo and Regulators
NHTSA will scrutinize Waymo’s technical data, records of incidents, and notes on software updates to evaluate whether the risk is systemic. If officials feel that there is still residual potential risk, they might seek further mitigations, operational limits in facilities close to schools, or a formal recall filing. Any recall would have to explain exactly how the software modifies stop-arms and flashing lights, and how the company tested its fix in actual on-the-road situations.
Operations may be driverless statewide in the great state of Texas, but operators of any motor vehicle are expected to obey traffic laws. That places the burden on Waymo to prove that its robotaxis can reliably interpret one of the most direct rules of the road: When a school bus is stopped with lights flashing and a stop-arm out, everybody stops. The result in Austin will mark how regulators believe autonomous fleets should navigate the highest-stakes, low-frequency situations that determine the public’s trust.
