Faulty AI face recognition is wrongly accusing alarming numbers of shoplifting

Facewatch uses facial recognition in participating UK retailers, matching shoppers’ facial geometry against a watchlist of people suspected of crimes. A possible match triggers an alert, but documented cases show staff treating alerts as grounds for confrontation, searches, removal, or bans. Big Brother Watch describes several innocent people being misidentified, including Danielle Horan, who was accused and removed from stores twice, and a teenager known as Sara, who was searched and told she was barred from other shops using the system. Facewatch later acknowledged that both its technology and a human reviewer had mistaken Sara for someone else. Because participating retailers share watchlist information, a false flag can follow someone across stores and spread their biometric data.

Responsibility is split: Facewatch says it is accountable for its technology and processes, while retailers are responsible for staff conduct. That division leaves wrongly accused customers with no clear remedy. UK data-protection law lets people request access to their personal data, usually with a response due within a month, but advocates say the process is burdensome and can require victims to give more identifying information to the organization that flagged them. The Information Commissioner’s Office found no need for regulatory action against Facewatch, while recommending audits and ongoing monitoring. The central argument is that claims of successful operation need independent evidence: retailers should publish alert volumes and false-match rates, watchlist sharing needs stricter limits, and staff must treat automated alerts as leads to verify - not proof of wrongdoing.