Facial recognition in stores: Shoppers wrongly tied to crime

A woman in Tennessee spent six months in jail for a crime in North Dakota. She'd never been to North Dakota. Police say a facial recognition system tied her to a bank fraud case, and the match was wrong.

Maybe you've never thought about facial recognition

Maybe you've never thought about facial recognition. But if you've ever walked into a store, a camera has probably looked at your face. And in some shops, software now decides whether you look like someone who stole something. It's unsettling, and it's reasonable to feel that way.

Here's where things stand. According to a report from Futurism, stores in the U.K. are using A.I. cameras that flag shoppers as suspected thieves. Customers say employees followed them, searched them, and told them to leave. Some say it happened in front of friends or family. And when they tried to clear their names, they found no clear path to do it. So the question is simple. When a computer says "that's the person," who checks whether the computer is right?

Start with the U.S. case that set the pattern. The Federal Trade Commission said the drugstore chain Rite Aid secretly used facial recognition on its customers for nearly a decade. The system produced thousands of false matches. And according to the F.T.C., those errors showed up more often in stores in neighborhoods with large Black and Asian populations.

Think about what a false match means at the

Think about what a false match means at the checkout line. An employee gets an alert. A stranger gets treated like a thief. The shopper did nothing wrong, and has no idea why it's happening.

Why do those errors fall unevenly? Federal testing gives a clue. According to data from the National Institute of Standards and Technology, the top-performing system as of March of twenty twenty-five made far more mistakes on some groups than others. It produced about three hundred and fifty times as many false matches for older West African women as for middle-aged Eastern European men. Three hundred and fifty times. That's not a small gap. It means the same camera can be reliable for one shopper and a coin flip for the next.

And the stakes go beyond a store. Jalil Richardson of North Carolina is counted as the fourteenth known person wrongly arrested because of facial recognition. Fourteen people we know about.

The companies have a defense

Now, the companies have a defense. Retailers say they're facing a shoplifting crisis. One U.K. vendor, Facewatch, claims its system is accurate nearly every time. It says it sent more than fifty thousand alerts to stores in a single month. But look at what that means. Even a tiny error rate, spread across fifty thousand alerts, still lands on real people.

Here's the part that should stop you. The developers of this technology have warned, repeatedly, not to treat a match as the only reason to act against someone. Yet in stores, staff act on the alert right away. They follow. They search. They call the police.

For anyone who investigates theft or fraud, that's a process failure. A match should start a human review, not end one. For the rest of us, it means a stranger's glance at a screen can decide how you're treated in public.