Waymo Recalls Nearly 4,000 Robotaxis Over Freeway Bug
Waymo is recalling nearly 4,000 self-driving taxis after at least 13 drove into closed highway construction zones in Phoenix and San Francisco.
Evgenii Arsentev · PhDWaymo is recalling nearly 4,000 of its self-driving taxis after a software flaw caused at least 13 of them to drive into highway sections that were closed for construction. The company filed a voluntary recall with the National Highway Traffic Safety Administration and says a software fix is currently being developed. No injuries were reported.
The pattern is specific. The vehicles failed to recognize ramp-closure signs and construction-zone warnings on highways. According to the filing, the software was either prioritizing the avoidance of other freeway hazards or failing to recognize the work zone at all. Six of the incidents happened in Phoenix in April; seven happened around San Francisco in May. After a cluster of cars drove into active construction zones on May 18, Waymo pulled all of its robotaxis off highways the next day, and its safety board decided to issue the recall on June 8.
Why a software 'recall' is different
When a regular automaker recalls cars, it usually means a physical part needs replacing. Here the 'defect' is a behavior baked into the driving model — the system was systematically misreading a certain kind of road situation. Waymo said it 'identified an area of improvement regarding performance around freeway construction zones,' and the fix is an over-the-air software update rather than a trip to the shop. That's the new shape of car safety: the dangerous flaw and its repair both live in code.
What it means for you
If you ride in one of these, the reassuring detail is the order of events: Waymo proactively notified state and federal regulators and took the cars off highways before the recall was finalized, not after a crash forced its hand. The sobering detail is that this is Waymo's sixth recall since it started operating, and the trigger was a mundane, everyday thing — roadwork. Self-driving systems are genuinely good at the routine 99% and still stumble on the weird edge cases that a human driver handles without thinking.
My take: this is roughly what responsible deployment is supposed to look like — catch a pattern, report it, ground the fleet, patch the model. The thing to watch isn't whether autonomous cars make mistakes; they will. It's whether the companies running them keep surfacing those mistakes quickly instead of quietly. On that score, a voluntary recall with the incidents laid out is the encouraging version of bad news.
Don't let one recall scare you off or lull you into trusting the car blindly. If you ride a robotaxi, stay aware near construction and roadwork — exactly the situations these systems are still weakest at — and treat the company's willingness to report problems as the real safety signal, more than any glossy 'miles driven' stat.
Related guides

Author
Evgenii Arsentev
PhD · Chief Product Officer at a tech company
Want to actually build this?
Guides explain. The free course transforms — personalized, gamified, and built to get you shipping fast.
◉ Start the free courseSource: techcrunch.com