I want robotaxis to work. I want autonomous vehicles to be real, boring, useful infrastructure. I want fewer drunk drivers, fewer distracted drivers, more mobility for people who cannot drive, and less of the daily stupidity that comes from putting humans behind the wheel of two tons of steel.
But a city street is not a demo floor, and an ambulance is not a debugging tool.
That is the line Waymo and the rest of the robotaxi industry keep trying to blur. TechCrunch reported on July 28 that Rep. Kevin Mullin is pushing the AV Emergency Response Coordination Act after incidents where autonomous vehicles blocked fire trucks and ambulances, entered emergency scenes, and failed around basic warning signals like cones and flares. Calling the bill a Luddite tantrum dodges the point. It asks companies to prove that their cars can get out of the way when lives are on the line.
That should be the easiest standard in the world to defend.
The average case is not enough
Waymo has a real safety argument, and anyone writing about this honestly has to admit it. The Insurance Institute for Highway Safety found that Waymo’s driverless vehicles had 68 percent fewer crashes per mile than human drivers across San Francisco, Phoenix, Los Angeles, and Austin when researchers limited the comparison to crashes a person would typically report to police. The same study found lower rates for single vehicle crashes and injury crashes. That matters. NHTSA reported 39,254 traffic deaths in 2024, and pretending human driving is acceptable because a robotaxi got stuck makes no sense.
The problem is that “safer than the average human driver” does not answer the emergency response question. A car can be excellent at routine lane changes and still be dangerous when smoke, fire, police tape, stalled traffic, flashing lights, or a sudden road closure turns the street into something the training distribution did not handle well. NHTSA said exactly that in its July 8 letter to driverless AV developers: emergency scenes are not rare edge cases. The agency said it had identified “a clear pattern” of driverless AVs interfering with law enforcement and first responders, including vehicles blocking ambulances and firefighters or failing to respond to lights, flares, smoke, fire, and cones.
That is the whole argument. Public safety cannot be graded on a weighted average where the failures are written off because the quarterly chart looks good. If a robotaxi blocks a fire truck, the family waiting for help does not care that the fleet has a good crash rate.
First responders should not be Waymo’s backup fleet
TechCrunch’s earlier investigation found at least six recent cases where first responders had to take control of Waymo vehicles and move them during emergency situations. One involved an officer responding to a mass shooting. Another involved an Atlanta first responder disengaging a Waymo after it entered an active crime scene. In a Redwood City fire response, TechCrunch reported that Waymo called 911 after one of its robotaxis could not move itself out of the way, and a California Highway Patrol officer eventually drove it to a park and ride lot.
This is the part that makes me angry. The failure does not stay inside Waymo. It spills into the public system. Firefighters, police officers, and paramedics are not idle capacity for a private company’s fleet operations. They are not there to become unpaid roadside assistance because a car with no driver cannot figure out what a human responder needs it to do.
San Francisco officials have been saying this plainly. At a March 2026 hearing, Mary Ellen Carroll, the city’s emergency management director, said public safety workers were becoming a default roadside assistance service for stuck Waymos, which she called untenable. District Supervisor Alan Wong put it even cleaner: first responders should not be AAA.
Waymo has built response machinery. Its own first responder guide describes a 24 hour hotline, remote assistance, an event response team, and roadside assistance. In February 2026, Waymo said it had about 70 remote assistance agents on duty worldwide for a fleet of about 3,000 vehicles. It said those workers advise the automated driving system rather than directly driving the cars.
Fine. Then the standard should be easy: if the system is ready for public scale, it should work when the city is under stress. If it still needs police officers to move the car, firefighters to clear the scene, and public officials to improvise around a stuck fleet, the private operating plan has not earned the public road space it is using.
The bill asks for basics, not miracles
Mullin’s bill is modest. It would require emergency response guides and rescue sheets. It would require companies to tell responders how to move, disable, tow, or immobilize a covered vehicle. It would require a hotline available at all times with a response within 30 seconds. It would create a geofence process so government officials can keep driverless vehicles out of emergency zones or safety hazards, with covered entities required to comply within two minutes.
That list is basic operational hygiene. Before companies put driverless cars on public roads, public servants need a reliable way to contact them, move their vehicles, recover data, reach passengers, and clear their fleets out of places where they can make an emergency worse.
California is already proving these ideas are not radical. In April, the California DMV adopted AV rules requiring 30 second first responder communication, manual override access, emergency geofencing, remote operations standards, and broader data reporting. The state also gave emergency officials power to create temporary zones that AVs cannot enter during public safety incidents, with vehicles already inside required to exit.
The companies should have asked for this themselves. A serious industry would want national emergency standards before expansion, because every ugly incident trains the public to see robotaxis as toys for rich neighborhoods and headaches for everyone else. Instead, the pattern has been familiar: deploy first, grow fast, tell cities the technology is inevitable, and negotiate safety after the public absorbs the cost.
Remote assistance is not a magic wand
The strongest defense is that autonomous vehicles need national rules, not a patchwork of city by city restrictions. That is true. It is also true that Waymo’s crash record looks better than human driving in important ways. I am not interested in banning useful technology because it is imperfect.
But useful technology has to survive contact with reality. The National Transportation Safety Board reported that on January 12, 2026, a Waymo in Austin stopped near a school bus that had its red lights flashing and stop arms extended. The vehicle asked remote assistance whether the bus had active signals. A remote assistance agent answered no. The Waymo then passed the bus while students were loading. NTSB said Austin ISD had reported multiple similar incidents, and Waymo later recalled 3,067 vehicles with its fifth generation automated driving system.
The incident does not doom robotaxis. It shows that “there is a human in the loop somewhere” should never be treated like a safety spell. A remote worker can be wrong. A process can be unclear. A hotline can be too slow. A vehicle can be technically autonomous and still dependent on human cleanup when the street gets weird.
NHTSA itself has acknowledged the regulatory gap. In a January 2026 speech, the agency said current rules lack automated driving system competency standards, even though vehicles that meet applicable federal motor vehicle standards can be deployed on public roads. The agency monitors crashes through its Standing General Order, and that system is limited: NHTSA says incident data is not normalized by fleet size, distance traveled, or operating domain, which means crash totals alone cannot tell the whole story.
That is why the public safety question needs its own rules. Crash rates are one metric. Emergency competence is another.
The street belongs to the public
Waymo told TechCrunch it supports a federal framework. Good. Then it should support enforceable standards that do more than bless what the companies already planned to do.
The future of transportation should be exciting. It should also be accountable. If a driverless car cannot understand that a fire truck needs the road, the answer is not to tell the public to be patient while the model improves. The answer is to keep that car out of the emergency zone until the company can prove it belongs there.
That is public accountability. It is the price of using public streets for private deployment. The roads were not built as a test track for Alphabet. They were built for people. Any robotaxi company that forgets that should learn the lesson before the sirens are trapped behind its product.