You’re driving down a two-lane road. Flashing lights appear in your rearview mirror. Instinct tells you to pull over—but you’re not driving. Your self-driving car sees the cop, but it also sees a solid double yellow line. So it stops dead in the middle of the lane. The officer gestures for you to move onto the shoulder—across that double line. Your car refuses. It’s following the law. Perfectly. And that’s exactly the problem.
We’ve been sold a simple story: autonomous vehicles will be safer because they obey traffic laws. No speeding, no rolling stops, no illegal U-turns. But what happens when the law itself demands flexibility? When a police officer, using hand signals, tells you to do something the car’s codebook says is forbidden? The machine doesn’t have common sense—it has a rulebook. And rulebooks don’t negotiate with sirens.
The very thing we designed them to do—follow every rule—is exactly what breaks the system.
Tesla’s upcoming Cybercab, a two-seat vehicle without a steering wheel, is programmed to respond to police hand signals, according to a June document. Waymo’s fleet pulls over at the sight of flashing lights. But these are point solutions, patches on a deeper wound. Hand signals vary by officer, by city, by mood. A cop’s job is to interpret the situation, wave you past a blocked intersection, or order you over a curb. That’s discretion—a human trait that no line of code can replicate.
In San Francisco, Waymo vehicles have already caused chaos by stopping at intersections where traffic directors waved them forward. The car saw a red light and froze. The officer saw a green hand and expected compliance. The result? Gridlock, frustration, and a growing sense that autonomous vehicles are polite to the point of paralysis.
Traffic laws are not a deterministic code; they are a set of guidelines for human interpretation, and police authority depends on human discretion—a power machines cannot contextually understand.
We assume the risk of driverless cars is that they’ll break the law. The real risk is that they’ll follow it so literally that the entire system of human policing breaks down. Every hand signal becomes a translation problem. Every unexpected road closure becomes a deadlock. And every officer is left wondering whether the car will obey them or their own programming.
This isn’t a minor bug—it’s a design philosophy failure. We’ve built vehicles that treat the law as a closed set of commands, when in reality, traffic enforcement is a live, human performance. The police don’t just enforce rules; they create temporary exceptions. And exceptions are exactly what machines flee from.
So the next time you see a driverless car, don’t worry that it’ll run a red. Worry that it’ll be too well-behaved to let you get out of the way.
FAQ
Q: Can't autonomous cars simply be programmed to obey police signals?
A: In theory, yes, but hand signals are not standardized, and officers use non-verbal cues (eye contact, gestures, urgency) that machines can't parse. Current solutions are vendor-specific and brittle—they work only for pre-defined scenarios, not the infinite variety of real-world police commands.
Q: What's the practical implication for cities rolling out AVs?
A: Cities need to rethink traffic enforcement for an era where vehicles cannot interpret human discretion. This may mean redesigning intersections, requiring AVs to connect to police dispatchers, or training officers to use standardized signals visible to sensors. Without that, expect public frustration and safety risks as machines and humans clash.
Q: But isn't perfect rule-following better than human error?
A: Only if the rules are deterministic. They aren't. Traffic laws rely on human judgment—police wave cars through reds, allow crossing double lines for emergencies, and improvise in real time. A car that cannot make those judgment calls creates a safety hazard. Perfect obedience without context is not safety; it's rigid inflexibility.