Your Face Was Scanned Saturday. Nobody Asked If That Was Legal.

Last month in Perth, Australia, a police van drove down public streets and scanned the faces of everyone who walked past. In one week, it captured more than a hundred and thirty thousand faces. And the one government official whose entire job is to protect people's privacy? She was never even asked.

If you've ever walked through a shopping center or

If you've ever walked through a shopping center or a public event, this is about you. You don't opt in to a face scan. You just walk by, and the camera does the rest. I know that feels unsettling — the idea that your face becomes data without anyone asking. But the scariest part of this story isn't the technology. It's what didn't happen before the van rolled out. So let me show you the invisible part — the safety checks that were supposed to happen, and why skipping them changes everything.

The Western Australia Police Force launched this trial with tech support from a company called N.E.C. The system worked like this. Cameras scanned every passing face. It compared each one, in real time, against a watchlist of about four thousand people with outstanding warrants. If your face didn't match, your image was pixelated and deleted almost immediately. If it did match, an officer nearby got an alert. Sounds reasonable, right?

Here's the problem. The person who should have reviewed all of this first — the state's Information Commissioner — was never invited to the table. No formal consultation. No review of how the data gets stored or destroyed. Nothing, before the first face was scanned.

Why does that matter so much

Why does that matter so much? Because of the accuracy gap that almost nobody talks about. When vendors brag about their rankings from N.I.S.T. — that's the U.S. government's testing lab — those scores come from perfect conditions. Frontal photos. Good lighting. High resolution. But researchers at Carnegie Mellon found that when a face turns to an angle, the algorithm's confidence can drop by thirty to forty percent. A street is nothing like a lab. People walk sideways, in bad light, in motion. That advertised accuracy quietly falls off a cliff.

And it gets worse for some people more than others. According to N.I.S.T.'s testing across different demographic groups, false matches can happen over a hundred times more often for the worst-performing group than the best. Part of that is lighting. Darker skin often needs different lighting to show facial structure. On an uncontrolled street, that means more false alarms for certain faces. For an everyday person, that's the difference between walking home and getting stopped by police for a match that was never really you.

Now let me correct the thing most people believe. You'll hear that these systems are about ninety percent accurate, and think — that's fine, an officer makes the final call anyway. That logic feels safe because ninety percent sounds like nine times out of ten. But that number only holds in perfect conditions. Out in the field, the algorithm still spits out a confidence score — say, eighty-seven percent. The officer sees that number and trusts it. They can't see that eighty-seven percent on a blurry sidewalk photo might mean a coin-flip in reality. The human becomes a rubber stamp for an error they can't detect.