UK Shop Face Recognition: Innocent Shoppers Named as Thieves

Shoppers in the U.K. have been walking into stores, getting stopped at the door, and told they're thieves. Some of them were shaking from head to toe afterward. They hadn't stolen anything. A camera said their face matched someone on a list.

If you've ever walked into a shop with a camera

If you've ever walked into a shop with a camera above the door, this story is about you. Most of us never agreed to be scanned. We just walked in to buy milk. Being pulled aside in front of strangers, for something you didn't do, is a frightening thing. That fear makes sense.

Some of the people wrongly flagged met in Parliament to tell their stories. According to TechRadar, they described the process of clearing their names as humiliating. Several called it Orwellian. At one South London Sainsbury's, a shopper was removed after the system misidentified them. There was no real way to opt out, and security staff knew little about how the system worked. So the question running through this episode is simple. When the camera gets it wrong, who fixes it?

Start with what happens after a wrong match. According to the people who testified, there's essentially no way to learn why you were flagged. There's no clear way to appeal. Once a name goes on a watchlist, it stays there for a year. And to get off it, victims say they had to hand over personal identification to a third party. So you prove you're innocent by giving more of your data to the company that flagged you. A mistake like that stops being a mistake. It becomes a sentence.

The company's side

Now the company's side. The firm behind British retail systems says its technology is accurate nearly every time. It says these cases are extremely rare, out of more than half a million alerts every year. It also blames human error in the stores, not the software. Maybe so. But look at what that argument admits. If the failure happens at the shop door, then the shop door is where the system breaks. For the person being stopped, it makes no difference whether the fault was the code or the cashier.

That's the heart of it. A face match is supposed to be a tip. Done right, the system compares a face to a list, flags a possible match, and asks a person to check. But some shoppers say they were stopped the moment they walked in. That suggests staff skipped the checking step entirely. A tip got treated like proof. And for a shopper, that means one wrong guess can ruin your afternoon, or your week.

We've seen this pattern in the United States. The F.T.C. forced a major pharmacy chain to stop using facial recognition. The system had produced thousands of false alerts. Employees followed customers, searched them, and accused them in public. The false alerts happened more often in neighborhoods with large Black and Asian populations. Research from N.I.S.T., the federal standards agency, backs up that pattern. Its testing found error rates vary a lot from one group of people to another. So "rare" doesn't land evenly. Some shoppers carry far more of the risk than others.