As of September 2026, Waymo remains under U.S. safety scrutiny after a string of investigations and software recalls. The most sensitive thread is NHTSA probe PE26001 opened after a January 23 Santa Monica incident: a driverless Waymo Jaguar struck a nine-year-old near a school; the child sustained minor injuries. NHTSA is examining school-zone caution, speed, vulnerable pedestrians, and post-impact response; the NTSB maintains a parallel investigation (HWY26FH008).
Waymo handed documents to the regulator, but much of the public response appears redacted, according to TechCrunch and August coverage. That opacity fuels political criticism: when a robotaxi interacts with children in a school setting, society demands auditable telemetry, not only press statements. Waymo says it cooperates and that aggregate safety rates remain competitive versus human drivers—a statistical argument that does not close the individual case.
Separately, software recalls—distinct from the child collision—have affected thousands of vehicles. Recall 25E084 covered about 3,067 systems whose ADS could proceed past stopped school buses with red lights or extended stop arms; updates completed in 2025. Recall 26E026 reached about 3,791 fifth/sixth-generation vehicles after an unoccupied robotaxi entered a flooded San Antonio roadway and was swept away; interim weather and mapping restrictions were applied.
Police and municipal interactions also weigh on the 2026 narrative. Local San Antonio reporting documented dozens of Waymo-related incidents—blockages, minor crashes, high water, and unusual public interactions—without each implying an ADS failure of equal severity. The nuance matters: volume of police contacts ≠ serious-injury rate, but it does measure social friction of driverless deployment.
The industrial context is robotaxi expansion across U.S. cities and competition with Cruise (after its crisis) and Zoox/Tesla. Each school or flood incident amplifies because the public calibrates trust in systems with no «face» to yell at. Regulators face the dilemma of not smothering innovation while not normalizing risk in vulnerable settings (children, people with reduced mobility, extreme weather).
Technically, the recalls show distinct failure classes: perception/policy regarding school-bus signals; planning in non-traversable water; possible remote-assistance errors in other NTSB threads. It is not a single bug. That complicates «already fixed» messaging and suggests continuous scenario validation, not only a generic urban ODD.
For Alphabet investors, the risk is social license more than CAPEX: a municipal veto or broad federal recall can delay profitable geographies. For insurers and cities, the question is how to allocate liability among ADS maker, service operator, and—if applicable—teleoperator. The Santa Monica child case will be a symbolic precedent even though injuries were minor.
Numbers should not be invented: as of September coverage, PE26001 remained open; consulted sources do not show a final NHTSA conclusion declaring a complete public root cause. Responsible journalism distinguishes «ongoing investigation» from «proven culpability.» Waymo denies patterns of systemic unsafety; critics demand more data disaggregated by vulnerable road-user type.
Internationally, European and Asian regulators watch the U.S. docket before authorizing fleets. A history of school-bus and flood recalls does not automatically block know-how exports, but it raises the evidence bar. Waymo will need to show not only miles without serious crash, but explicit slow-down policies near schools and no-go rules on updated flood maps.
The lesson for the whole autonomous-driving industry is that perceived safety is fought in the distribution’s tail: the rare case beside a school outweighs a thousand quiet miles in headlines. Ethical instrumentation—road-facing cameras with privacy, immutable logs, third-party auditors—may matter as much as the next perception model.
In parallel, the general AI debate (slowdown, threat reports) rarely connects to robotaxis, yet both share governance of autonomous systems in the physical world. A chat agent that hallucinates and an ADS that misreads a stop arm are different failures; the civic demand for explainability and human emergency brakes is analogous.
In sum, in September 2026 Waymo faces NHTSA/NTSB scrutiny over the Santa Monica school incident (January), software recalls for school buses (~3,067) and flooding (~3,791), and local police friction. Assert carefully: open investigations, minor injuries in the child case, recalls separate from the crash. Neural Pulse will follow PE26001 resolutions and ODD changes.
Cities negotiating MOUs with robotaxi operators can demand post-incident transparency clauses: log-delivery deadlines, independent researcher access, and operational-pause thresholds after crashes involving vulnerable users. Waiting only on NHTSA may be too slow for local trust. Santa Monica and San Antonio already offer municipal-governance lessons.
Waymo, for its part, wins if it publishes more disaggregated data—not only favorable averages—and if it demonstrates measurable improvements in school zones and flood maps. Robotaxi safety will be won in the distribution’s tail and in public perception, not only in accumulated miles. September 2026 leaves the docket open; the next chapter will be written by PE26001 and residual recalls.
The recalls show Waymo can patch entire fleets via software; that same capability should be used proactively when telemetry shows school-zone near-misses, not only after an ODI. The AV industry will mature when over-the-air updates are a culture of prevention, not only post-headline correction.
