Robotaxis Can Get Flagged, But Not Really Fined: The Accountability Gap in Autonomous Systems
California has started to close one of the stranger gaps in the robotaxi era: what happens when a driverless car breaks a traffic rule and there is no driver to ticket.
The Mercury News reported May 1 that new California Department of Motor Vehicles regulations, issued April 28 and taking effect July 1, will let police issue notices of “noncompliance” when autonomous vehicles commit moving violations. An officer can place the notice in the vehicle or hand it to a company representative. The company must then report the incident to the DMV within 72 hours, or within 24 hours if the officer marks it as a priority because there was a clear or potential danger or risk of injury.
That is progress. It creates a record. It tells the company and the regulator that something happened. It gives police a formal channel instead of leaving them to wave at a machine that cannot roll down a window, explain itself, or sign a citation.
But the gap is still obvious. The Mercury News reported that, unlike human drivers, robotaxi companies do not face ordinary traffic fines or license-style points under the new process. The DMV is using its administrative authority over autonomous-vehicle permits rather than turning a software-operated vehicle into a normal traffic defendant. Futurism, summarizing the same California shift, also noted a separate emergency-response rule: autonomous vehicles must comply with emergency personnel within 30 seconds of an instruction, with permit consequences available if companies fail the standard.
So the real issue is not whether California is doing nothing. It is not. The real issue is that the first layer of accountability for physical AI is being built as an administrative reporting system, while the street-level public experience still runs through rules designed for human drivers.
That mismatch is the story.
The ticket was never just about the money
A traffic ticket looks small. It is a fine, a form, a court date, a point on a record, a threat to a license or insurance rate. Most people experience it as a personal penalty.
But in political terms, a traffic ticket is also a compact accountability machine. It identifies a person. It creates a record. It imposes a cost. It gives the government a way to say: if you use public roads, you are subject to public rules.
That system assumes a human at the center. The driver can be stopped. The driver can be cited. The driver can be fined. The driver can lose points. The driver can have insurance consequences. The driver can contest the ticket in a process built around individual responsibility.
A robotaxi breaks that model. The passenger is not driving. The vehicle is not a legal person in any ordinary civic sense. The software cannot appear in traffic court. The remote support team may not have made the decision in the moment. The company owns or operates the system, but the violation happens through a machine that turns law into sensor inputs, maps, models, and commands.
That is why this California rule matters. It is not just about Waymo, one intersection, or one weird traffic incident. It is a preview of what happens when law written for people meets machines that act in public space.
The state can document the behavior. It can require reporting. It can threaten permit consequences. But the old penalty chain does not snap neatly into place.
A note home from the teacher is not the same as a ticket
Carnegie Mellon professor emeritus Phil Koopman, who studies autonomous vehicles, told The Mercury News the DMV was trying to make progress based on problems that have already gone wrong. He also compared the notice system to “a note sent home from the teacher,” while asking how much a $300 fine would matter to a company like Waymo anyway.
That is the right tension. A normal fine might be too small to discipline a large autonomous-vehicle company. But the absence of ordinary fines still matters because it changes the relationship between local enforcement, corporate deployment, and the public.
If a human driver blocks an intersection, ignores an officer, passes a stopped school bus, or creates a hazard, the public has a familiar accountability script. Maybe the ticket is not enough. Maybe the fine is too low. Maybe enforcement is inconsistent. But the rulebook has a visible consequence.
With robotaxis, the public may see something different: the vehicle breaks a rule, police file a notice, the company reports to the DMV, the regulator adds it to a compliance file, and whatever consequence follows is slower, more technical, and less transparent.
That may be rational from an administrative standpoint. Regulators can look at patterns instead of isolated events. They can suspend or condition permits if a company repeatedly creates problems. They can target the operating authority that matters most to the business.
But citizens experience streets locally. They do not live inside DMV compliance dashboards. They live with blocked lanes, confused emergency scenes, delayed buses, near misses, and the ordinary expectation that the rules apply equally to everyone using the road.
If the public sees human drivers punished immediately while software fleets move through an administrative back channel, trust will not improve. It will erode.
The incentive problem is bigger than one company
The incentive question is simple: what happens if the upside of deployment is private and the cost of enforcement is public?
Autonomous-vehicle companies have powerful reasons to scale. More miles produce more data. More data improves systems. More presence normalizes the product. More geographic coverage increases investor confidence and commercial value. The industry’s logic rewards getting onto roads, learning from real environments, and expanding from city to city.
Local governments have a different problem. They must manage the public right of way. They must protect emergency access. They must keep buses, bikes, pedestrians, construction crews, school zones, and human drivers moving through the same streets. When a driverless vehicle behaves badly, police and emergency workers absorb the first cost.
South Bay state Sen. Dave Cortese, chair of the state Senate Transportation Committee, told The Mercury News that the regulations burden local jurisdictions with enforcing traffic laws on robotaxis without the fine revenue that helps pay for enforcement against human drivers. That is not a minor detail. It is the political economy of automation in miniature.
The company gets a deployment path. The state gets a regulatory file. The city gets the street-level headache. The police officer gets the awkward job of serving paperwork on a machine. The public gets asked to believe that the system is under control.
This is how accountability gaps usually appear. Not with a villain announcing that laws no longer matter. With a new technology landing inside an old enforcement model and everyone discovering, after the fact, that the incentives no longer line up.
Administrative penalties may matter — but only if citizens can see them
It would be a mistake to say robotaxi companies face no possible consequences. California’s DMV can regulate autonomous-vehicle permits. Futurism noted that emergency-response compliance can connect to permit suspension. The Mercury News described the noncompliance notices as part of the DMV’s administrative authority over autonomous vehicles.
That matters. Permit authority may be more serious than a small traffic fine if regulators use it aggressively and transparently. A fleet operator probably cares more about losing permission to operate than about paying isolated citations.
But that only works if the system becomes legible.
How many notices were issued? For what conduct? In what cities? How quickly did companies report them? Did the DMV treat them as isolated glitches or signs of a pattern? Did any permit condition change? Did any company face a meaningful operational consequence? Were emergency-scene failures treated differently from low-level moving violations?
Those are not anti-technology questions. They are basic public-accountability questions.
If administrative enforcement happens behind a curtain, the public will read the system as indulgence. If it happens in a public, standardized, searchable way, it can become a real accountability layer. The difference is not the existence of a form. The difference is whether the form changes behavior.
Physical AI needs physical accountability
We have spent years talking about artificial intelligence as if it lives in screens. Chatbots. Search summaries. Recommendation feeds. Ad systems. Automated decisions in hiring, lending, insurance, and benefits.
Robotaxis are different because they put software directly into physical public space. The model is not just producing text or ranking content. It is navigating roads, responding to police, interpreting school buses, reacting to construction zones, and sharing streets with people who did not consent to become part of a live deployment environment.
That raises a broader question for the next decade: when autonomous systems break public rules, who is accountable?
The vehicle? The model? The company? The fleet operator? The passenger? The remote-support staff? The vendor that supplied the mapping data? The state agency that approved the permit? The city that allowed expansion? Nobody?
Every autonomous system will try to route that question through a slightly different legal structure. Drones will raise airspace and privacy questions. Delivery bots will raise sidewalk and disability-access questions. Autonomous trucks will raise highway and labor questions. Warehouse robots will raise workplace-safety questions. AI agents operating inside businesses and government agencies will raise contracting, fraud, discrimination, and due-process questions.
The pattern will repeat: a machine does something consequential, but the enforcement system was built around a human actor.
That is why the robotaxi ticket problem is not a curiosity. It is a template.
The wrong answer is either panic or permissionless immunity
There are two lazy ways to discuss autonomous systems.
One is panic: every robot is unsafe, every company is reckless, and every deployment is a dystopian plot. That is not serious. Autonomous technology can reduce some risks, improve mobility for some people, and eventually make parts of transportation more efficient. The public does not need theatrical fear.
The other lazy answer is permissionless immunity: because the technology is innovative, regulators should let companies deploy first and solve the accountability problems later. That is also not serious. Public roads are not private laboratories. A city street is a shared civic asset, not a beta-testing surface with pedestrians as background variables.
The serious answer is boring but hard: build the accountability chain before scale makes the behavior politically untouchable.
That means regulators need more than incident reporting. They need clear escalation ladders. They need public dashboards. They need local reimbursement when enforcement costs are pushed onto cities. They need rules for emergency compliance that are not merely aspirational. They need to define when repeated software behavior becomes a permit problem, not a customer-service problem.
Companies should want this too, if they are thinking beyond the next expansion map. Public trust is not created by insisting the technology is statistically safe. It is created when people believe the system will tell the truth when something goes wrong, impose real consequences when patterns appear, and treat public rules as obligations rather than obstacles.
The politics of the next machine age will be enforcement
The robotaxi story exposes a broader political reality: the next phase of AI will not be judged only by capability. It will be judged by enforceability.
Can a rule be applied to a system that has no driver? Can a citizen challenge an automated decision that has no obvious human owner? Can a city recover costs when a company’s machine creates public burdens? Can a regulator punish repeated software behavior without waiting for a catastrophe? Can the law distinguish a one-off glitch from a business model that quietly externalizes risk?
These questions sound technical. They are political.
Power often moves fastest when rules are ambiguous. Companies do not need formal immunity if the practical penalty system does not know how to reach them. They do not need to defeat traffic law if traffic law cannot decide who the offender is. They do not need to abolish accountability if accountability gets converted into a compliance memo that ordinary citizens never see.
California’s new rule is a start. It acknowledges that police need a way to flag robotaxi violations and that autonomous-vehicle companies must report those incidents to the DMV. It also shows how much work remains. A notice is not the same as a ticket. Reporting is not the same as punishment. Permit authority is not the same as public accountability unless it is used clearly and visibly.
The public should not ask for a world where every new machine is blocked by old rules. But it should also reject a world where old rules are used as an excuse for weak consequences.
The right standard is simple: if a company puts software into public space, the accountability has to be as real as the deployment. Not symbolic. Not delayed until after trust collapses. Not hidden in a regulator’s file.
Real.
That is the test robotaxis are bringing to California streets. And it is the test every autonomous system will bring next.