Sainsbury's Pauses AI Face Scanning After Wrongly Accusing Shopper of Theft
A comedy promoter was publicly ejected from a London Sainsbury's after facial recognition software flagged him as a shoplifter — the latest in a run of similar mistakes at UK retailers.
What happened: Matt Arnold, a 46-year-old comedy promoter, was buying supplies at a Sainsbury's in East Dulwich, London, when two managers stopped him mid-checkout, told him he couldn't be served over an earlier incident, and tried to walk him out of the store. Leaving, he saw his face circled in red on an overhead CCTV alert screen. The store uses Facewatch, an AI facial recognition system meant to flag known shoplifters. Sainsbury's apologised the next day and paused Facewatch at that one store while it investigates, but says the software wasn't the problem, it was mishandled by staff.
Why it matters: This isn't a one-off. A Sainsbury's shopper in Elephant and Castle was wrongly flagged by the same software last year, and similar misidentifications have hit Home Bargains and B&M stores. Arnold says he stayed calm and pushed back, but worries about what happens when someone more vulnerable, or younger, is wrongly accused in public and doesn't have the confidence to resist being marched out. He's also unsettled that staff followed the alert without pausing to notice he wasn't behaving like someone fleeing a theft.
How it works, plainly: Facewatch scans shoppers' faces against a shared watchlist of people flagged across participating stores. When it finds a match, an alert pops up on staff devices and stays visible for up to an hour, meant to be checked by a trained manager before anyone acts. Sainsbury's says the system is 99.98% accurate and every alert is logged. But that still leaves room for real, repeated mistakes once you factor in how staff interpret and act on the alert in the moment, which is exactly what both Sainsbury's and Facewatch say went wrong here.
The fallout: Sainsbury's paused Facewatch only at the East Dulwich store, calling it a precaution while staff get extra training, not an admission the technology itself is flawed. Arnold wants it suspended everywhere, arguing that since it's a centralised system, a fault in one store could exist in all of them. Facewatch maintains its alert was correct and blames the in-store response. With no company-wide pause and no clear outside review announced, other shoppers face the same odds the next time an alert fires.