Zoox is recalling all 105 of its robotaxis after one drove into a smoke-obscured fire scene in June. It's part of a pattern regulators say autonomous vehicle developers still haven't solved.
On June 20, an unoccupied Zoox robotaxi drove into an active fire scene that was obscured by heavy smoke and hadn't yet been marked off with traffic cones. The vehicle braked hard, then reversed out of the area under guidance from a remote operator. No injuries were reported.
Zoox is recalling all 105 vehicles in its fleet as a result, and issued a software update on July 15 aimed at improving how the system recognizes heavy smoke, including smoke from EV fires.
There is a clear pattern of autonomous vehicles failing to properly respond to emergency scenes.
🎙️ Want the full conversation?
Listen to today's Automotive State of the Union episode for the complete discussion, additional context, and the conversations that shaped our perspective.
What Actually Happened on June 20
The vehicle wasn't carrying a passenger, and the incident resolved without harm. But it's a clean, specific example of exactly the kind of scenario regulators have been warning about: an AV encountering a real-world hazard that a human driver would recognize instantly, and struggling to interpret it correctly.
This Isn't a One-Off. NHTSA Says It's a Pattern.
Morrison's comments came just days after NHTSA sent letters to autonomous vehicle developers broadly, citing a pattern of AVs interfering with emergency responders, failing to recognize flashing lights, flares, smoke, or cones, and in some cases blocking ambulances outright. Zoox's recall isn't an isolated company problem. It's one data point in a wider regulatory concern spanning the whole AV industry.
That distinction matters if this comes up in a customer conversation. A single company issuing a fix doesn't mean the underlying challenge, teaching autonomous systems to recognize emergency scenes the way a human instinctively does, is solved industry-wide.
What This Means for Your Store
If AV trust and safety comes up with a customer, this is a useful, specific example to have ready rather than a vague reassurance either way. The real story here isn't "an AV failed." It's that a real incident got identified, addressed, and patched within about three weeks, and that the company disclosed it and issued a fix rather than waiting for it to happen again.
That's a more honest, more useful answer than either "AVs are dangerous" or "AVs are already perfect." Neither extreme matches what's actually happening in this space right now.
