You've probably seen this before. Waymo's driverless cars have roamed San Francisco and Phoenix for years — millions of paid rides with no one at the wheel. We've stopped finding it strange.

Here's what should still feel strange. In all those millions of driverless miles, we never answered the simplest question underneath them: when one of these cars hits someone, who's responsible? This month, Tesla made it urgent — putting cars on Miami streets with no safety driver from day one, camera-only, straight into an open federal safety probe. The technology races ahead. The answer stays blank.

For a year, Tesla did this carefully — a human safety monitor rode along in Austin, Dallas and Houston for months. Miami skipped it: fully driverless from day one, on cameras alone, no lidar or radar like Waymo's cars carry. The service is fenced to about a dozen square miles, avoiding downtown and the airport.

Why Miami? Florida requires no permit for a driverless car. The enabling law came from next door — a Texas rule took effect May 28th, and the same day Tesla signed its own paperwork certifying Level 4, no human oversight. Meanwhile, federal regulators have been probing this exact camera-only system since March, after finding it can fail to see and warn in poor visibility. The investigation is open. The cars are already driving.

The easy fight is whether these cars are safe — maybe even safer than us. That debate will run for years, and it misses something. Because safe or not, sometimes the car will crash. And in that moment a whole invisible system is supposed to switch on — fault, punishment, payment, deterrence — every part of it assuming a human was driving. So the real question is quieter: when you remove the driver, where does all that accountability go?

When a human crashes, a loop closes on them: fault, a ticket, points, a higher premium — a signal, felt that week, nudging millions to drive a little better. Take the wheel away and every arrow that pointed at the driver now points at an empty seat. You can't ticket a car or send software to court.

And who can actually make the next car safer? Only the company that writes the software. But nothing gives it the driver's fast, per-crash jolt — there's no ticket for a line of code. The one lever that reaches the maker is a regulator's recall: slow, rare, reactive. The constant pressure that keeps human drivers careful barely touches the one company that could make the car safe.

So who should the arrow point at? Point it at the maker — treat the car like any product, where the company is liable, not the passenger; Britain wrote exactly that into law in 2024. Or stop fighting about fault: a no-fault pool, funded per mile, pays anyone hurt fast, the way New Zealand runs its whole injury system. Or borrow from aviation: certify the system before launch, and have an independent body with no stake dissect every crash.

Each pays a price — slower deployment, a dulled safety signal, or heavy bureaucracy. But none of them argues whether the car is safe. They assume it sometimes won't be, and decide in advance who answers. That's the part we're leaving blank.

The car pulling away from the curb is the easy part to notice. What you can't see is everything underneath it — a hundred years of rules assuming a person was in charge. We're removing that person, and not just from cars: software already decides who gets a loan, who gets hired, how markets trade. Each time, the human the rules were built to hold walks out of the seat. So the question isn't whether the machine is better than us. It's: when it fails — and it will — who's still in the seat? If the honest answer is no one, we didn't automate the driving. We automated away the blame.