Sainsbury's Pauses Facial Recognition After Wrongly Accusing a Paying Customer

Sainsbury's has paused its Facewatch AI-assisted facial recognition system at its East Dulwich store after the technology wrongly flagged paying customer Matt Arnold, who was asked to leave on August 6. The retailer attributes the incident to human error, but Arnold's case follows a similar false accusation in January involving Warren Rajah, raising concerns about the system's deployment across 55 stores.

An AI camera meant to help Sainsbury's spot repeat offenders instead ended with a paying customer being marched out of a south London store. The company calls it human error, but that excuse gets weaker each time it happens. Matt Arnold had scanned his groceries and tapped his Nectar card at Sainsbury's East Dulwich superstore on August 6 when two managers came over and told him he couldn't be served. He looked up. An overhead CCTV monitor showed his own face circled in red. The 46-year-old comedy promoter was buying drinks and snacks for a standup night at nearby Dulwich Hamlet Football Club. Then he was asked to leave. According to The Guardian, Sainsbury's has paused its Facewatch AI-assisted facial recognition system at the East Dulwich store while it investigates. The retailer says the software wasn't at fault. A Sainsbury's spokesperson said the incident was caused by human error, not the facial recognition technology, and said every match is meant to be reviewed by a trained manager. Facewatch gave the same basic account: a correct alert was sent to the retailer, but it was handled wrongly in the store. You can call that human error. Arnold calls it humiliating. He said he felt embarrassed and mortified - powerless, too. He also asked the obvious question: if Sainsbury's is worried enough to pause the system in East Dulwich, why keep it running everywhere else? His concern isn't only about himself. It's about the shopper who doesn't argue back when a manager says they're not welcome: the teenager, the anxious customer, the person who just leaves because the screen and the staff have already made the accusation feel official. Researchers Show How Microsoft Copilot Can Turn One Login Into $247,500 https://startupfortune.com/researchers-show-how-microsoft-copilot-can-turn-one-login-into-247500/ Barracuda Networks built a proof of concept showing how a single compromised Microsoft 365 login can be escalated, via the account's own Copilot assistant, into a $247,500 wire fraud. Researchers call the technique living off the AI land, or LOTAIL, because it uses no malware, only the AI's legitimate access to search, draft and delete emails. - how Microsoft Copilot can be exploited for fraud https://startupfortune.com/researchers-show-how-microsoft-copilot-can-turn-one-login-into-247500/ - stolen employee login wire fraud security vulnerability https://startupfortune.com/researchers-show-how-microsoft-copilot-can-turn-one-login-into-247500/ Not the first wrong customer Arnold's case follows another public failure at Sainsbury's this year. Warren Rajah, a 42-year-old data strategist from Elephant and Castle, said he was told to leave his local branch on January 27 after staff mistook him for an offender flagged by the Facewatch system. The Press Association reported that Rajah later contacted Facewatch, which told him he wasn't on its database after he sent a passport copy and a photo of himself. Sainsbury's apologised and offered him a £75 shopping voucher. That safeguard failed. Facewatch scans shoppers against a watchlist built from past incidents: theft, violence, aggression. The company and Sainsbury's both say a person reviews every alert before any action is taken. That review step is meant to be the protection between an automated match and a public accusation. In both Sainsbury's cases now in the public record, the decisive failure happened after the alert, when staff dealt with the wrong person. Frankly, that distinction only helps so much. If a system depends on hurried store workers interpreting alerts correctly in the middle of a shop floor, then the human part is part of the system. You don't get to separate the software from the moment it sends someone toward a customer with authority in their voice. The rollout is still growing The stakes are bigger than one East Dulwich store. Retail Gazette reported in July that Sainsbury's already had the AI-powered Facewatch system live in 55 stores and planned to add it to as many as 150 more supermarkets before Christmas. The Grocer reported the same expansion as a move toward more than 200 stores by the end of the year. Sainsbury's has said its early trial saw incidents fall 46% - theft, harm, aggression, antisocial behaviour - while 92% of identified offenders did not return. Retailers have real reasons to want this tool. Shop workers deal with abuse and theft, and supermarkets are under pressure to protect staff without putting a guard at every door. But every wrong approach carries a cost too. It falls on the customer standing at the till. Paid shopping, a loyalty card, a face suddenly framed as evidence. Sainsbury's says the East Dulwich system stays off while staff get more training and the incident is investigated. No public timeline. Nothing on when, or even whether, the system comes back to that branch. The rest of the rollout is the harder question: training is exactly the control Sainsbury's already said was in place. Micron Launches a $250 Million Fund to Bankroll the AI Startups Buying Its Chips https://startupfortune.com/micron-launches-a-250-million-fund-to-bankroll-the-ai-startups-buying-its-chips/ Micron launched a $250 million venture fund, the Micron Ventures Paradigm Fund, to invest in AI startups across model architecture, compute, enterprise applications and physical AI. It's the company's third and largest fund, following similar strategic-investment moves from SK Hynix and Nvidia. - micron ventures paradigm fund for AI startups https://startupfortune.com/micron-launches-a-250-million-fund-to-bankroll-the-ai-startups-buying-its-chips/ - how chipmakers invest in AI infrastructure companies https://startupfortune.com/micron-launches-a-250-million-fund-to-bankroll-the-ai-startups-buying-its-chips/ Twice is a pattern. A company can say the camera worked and the person failed. But you, as a shopper, experience the whole chain as one decision: the camera, the alert, the manager, the public walk out of the store. If Sainsbury's wants to expand Facewatch to hundreds of locations, it has to prove the weakest step can hold under ordinary pressure. Scale is the problem. Also read: Groq Raises Another $350 Million to Build Out Its Nvidia-Powered Neocloud https://startupfortune.com/groq-raises-another-350-million-to-build-out-its-nvidia-powered-neocloud/ • Amazon Is Cutting the Spines Off Rare Books to Train Its AI Models https://startupfortune.com/amazon-is-cutting-the-spines-off-rare-books-to-train-its-ai-models/ • Researchers Show How Microsoft Copilot Can Turn One Login Into $247,500 https://startupfortune.com/researchers-show-how-microsoft-copilot-can-turn-one-login-into-247500/