Alarm over launch of facial recognition in UK shops that instantly alerts police
Facewatch, a facial recognition system already used by more than 100 UK businesses including Sainsbury's, B&M and Spar to monitor suspected thieves, is launching a feature that alerts police in real time when its cameras identify serious offenders. The company says the "UK-first" development, due in autumn, will notify police in an average of four seconds when the "worst offenders" trigger a live match. Civil liberties groups have condemned the move as a "dangerous escalation" towards surveillance and criminalisation in the retail sector, arguing the technology has "shot on far ahead of the regulation".
Critics from Liberty, Open Rights Group, the Ada Lovelace Institute and Big Brother Watch warn that the system is untested and opaque, that it risks summoning police over people who have committed no crime, and that it makes mistakes — with evidence suggesting black and Asian people are more likely to be misidentified. Facewatch says it alerted retailers almost 300,000 times about "known repeat offenders" in the first half of 2026, against a backdrop of 509,566 recorded shoplifting offences in England and Wales in the year to December 2025. Concerns were also raised that the government's planned legal framework for facial recognition would not cover the private sector, creating a "backdoor" partnership with police that is not held to the same standards; use of the technology is set to expand, with Sainsbury's increasing deployment from 55 stores to more than 200 by year's end.
New here? Start with this
Facewatch is a facial recognition system used in more than 100 UK retail shops, including major chains such as Sainsbury's and B&M, to identify and monitor people suspected of repeat shoplifting. The technology has been deployed to address retail theft, which costs businesses significantly, with more than 500,000 recorded offences in England and Wales annually.
Civil liberties groups, including Liberty, Big Brother Watch and the Open Rights Group, have raised concerns about the reliability and regulation of facial recognition in the retail sector. Research suggests these systems misidentify people from black and Asian backgrounds at higher rates, and critics argue the technology is being used without sufficient safeguards or transparency about how data is handled.
A concern is that facial recognition in retail has developed faster than legal regulation, particularly in the private sector where oversight is limited. This raises questions about the accuracy of the technology, the privacy of people being monitored, and what protections exist if identifications are wrong.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters would argue that with shoplifting exceeding 500,000 recorded offences annually, retail businesses and communities face genuine security challenges warranting technological solutions. The system targets known repeat offenders rather than innocent shoppers, and real-time police alerts could improve response times to serious criminals, protecting staff and customers whilst helping police allocate resources more efficiently. Retailers have legitimate commercial interests in deploying security measures, and automated identification offers a more effective approach than manual monitoring.
The case against
Critics argue that facial recognition technology remains untested and opaque, with documented evidence of higher error rates affecting people of colour, creating genuine risks of wrongful police dispatch and harassment based on misidentification. The system operates outside planned legal frameworks, creating a backdoor surveillance partnership between police and the private sector that bypasses democratic accountability and civil liberties protections. Rapid expansion without proper safeguards or public consent raises fundamental concerns about state and corporate surveillance that outweigh security benefits, particularly when innocent people face police intervention due to algorithmic errors.