Driverless Waymo taxis under investigation for failing to stop for a school bus (www.latimes.com)

The NHTSA has opened a preliminary investigation into roughly 2,000 Waymo vehicles after a Sept. 22 incident in Atlanta in which a Waymo car using the company’s fifth‑generation automated driving system initially stopped for a school bus but then drove around the front and passed the left side while the bus’s stop arms and flashing red lights were active. Regulators say similar events have likely occurred; Waymo says it has deployed software updates, will push more, and that the vehicle kept a safe distance because the bus partially blocked a driveway. The probe follows other operational issues (including a 1,212‑vehicle recall and an observed illegal U‑turn in San Bruno), and highlights limits in enforcement because police currently cannot issue standard traffic citations to driverless cars. For the AI/ML community this raises technical and systemic questions about perception, decision‑making and safety validation in rare but high‑risk scenarios. Handling school buses requires robust detection of physical actuators (stop/crossing arms), intent inference for disembarking children, and conservative planning when road geometry is ambiguous (e.g., blocked driveways). The case underscores the need for scenario‑based testing, formal verification or interpretable failure modes, fleet telemetry for rapid patching, and stronger regulatory and legal frameworks for accountability. NHTSA oversight of a deployed fleet sets an important precedent for how autonomous systems must demonstrate compliance and safety under real‑world edge cases.