Security cameras have become an ordinary feature of modern shopping. They record entrances, tills and aisles, helping stores investigate theft, aggression and other incidents. Facial recognition changes the nature of that surveillance. Instead of simply recording what happens, software can analyse faces as people enter and compare them with a database of selected individuals.
Retailers argue that this can help protect staff, customers and stock. Critics warn that it can turn routine shopping into a form of biometric screening, often in places people cannot easily avoid. The question is no longer only whether a camera is watching, but whether a system is attempting to identify everyone who walks past it.

From recording incidents to identifying people
Conventional CCTV stores footage for later review. Facial recognition software can add an automated identification layer. A camera captures a face, software detects and measures facial features, and the resulting biometric template is compared with entries on a watchlist. The system may flag a possible match to a member of staff, who is expected to assess it before any action is taken.
That distinction matters. A retailer using ordinary CCTV is not necessarily identifying customers by name. A live facial recognition system may process the faces of large numbers of people to find a much smaller number of individuals already recorded on a watchlist. The majority of shoppers may have no connection with suspected wrongdoing, yet their faces can still be analysed during the process.
Watchlists may contain people suspected of repeated theft, individuals previously involved in violence or people banned from particular premises. The criteria for inclusion, quality of the evidence and process for challenging an entry are therefore crucial. A mistaken or outdated entry can follow someone into a shop where they have never caused trouble.
Why retailers are interested
Retail crime can impose substantial costs and can expose shop workers to threats and abuse. Retailers say facial recognition may alert staff when someone with a documented history of serious incidents enters a store, allowing a response before a confrontation develops. It may also help security teams identify repeat offenders across several branches more quickly than manual review of recordings.
The technology has been discussed and tested by major supermarket businesses in several countries. In August 2026, reporting on Australian supermarket trials described Coles and Woolworths exploring facial recognition, while both companies emphasised that their tests did not amount to a full customer-facing deployment. The debate illustrates the distinction between a limited technical trial and a decision to scan shoppers as a routine practice.
In Britain, some retailers and other private venues have already used commercial facial recognition services. A 2023 investigation by The Guardian described deployments at retail sites, including a garden centre in south-east London. Such examples demonstrate that the issue is not confined to future technology or hypothetical proposals.
The public-private surveillance question
One of the most difficult issues arises when private organisations and police cooperate. A shop may collect information for its own security purposes, while police have separate responsibilities and legal powers. When data, watchlists or live camera feeds are shared, the boundary between commercial security and public surveillance can become less clear.
The UK Biometrics and Forensics Ethics Group examined this issue in a briefing updated in May 2026. It considered collaboration between police and private organisations in locations such as shops, shopping centres and airports. The group highlighted questions about necessity, proportionality, accountability and the expectations of people moving through privately owned public spaces.
Private ownership does not remove data-protection obligations. The Information Commissioner’s Office says organisations must be able to explain why facial recognition is necessary, identify a lawful basis and a special-category condition where required, assess risks and consider whether less intrusive alternatives could achieve the same purpose.
Consent is harder than it sounds
A notice at a shop entrance may tell customers that facial recognition is in operation, but notice is not the same as freely given consent. People may need to enter a supermarket to buy food or use an essential service. If the only practical way to shop is to pass through a scanning zone, the ability to refuse may be limited.
The ICO’s guidance gives the example of a gym requiring members to use facial recognition with no alternative access method. In that situation, consent may not be freely given because a person cannot realistically decline and still use the service. The same principle is relevant to retail environments: a business should consider whether a less intrusive method, such as conventional CCTV, trained staff or targeted security measures, would be sufficient.
False matches and the consequences of suspicion
Facial recognition does not deliver certainty. It produces a similarity score and a possible match, based on image quality, camera position, lighting, facial appearance and the system’s threshold. A false alert can lead to an innocent customer being watched, challenged, searched or excluded. Even if staff later establish that the match was wrong, the experience can be humiliating and distressing.
Human review is essential, but it is not a magic safeguard. Staff need appropriate training, enough time and clear instructions not to treat an alert as proof. They should check the underlying evidence and avoid escalating a match solely because a computer has raised it. Systems should also be tested in real operating conditions and monitored for differences in performance across demographic groups.
What should a responsible system look like?
Any deployment should have a clearly defined purpose, a documented assessment of necessity and proportionality, a carefully controlled watchlist and a limited retention period. Customers should be told what is happening, who controls the data, what the system is checking and how to raise a concern. People should have a meaningful route to challenge an inaccurate listing or a mistaken intervention.
Independent testing and transparent reporting would help the public assess whether claimed security benefits are real. Retailers should publish enough information to explain how many alerts are generated, how often they are confirmed and what happens to data relating to people who are not on a watchlist. A system that cannot demonstrate value beyond ordinary security measures may not justify the intrusion.
A shop is not a border checkpoint
There is a difference between verifying a traveller’s identity at an international border and scanning customers who are simply going about everyday life. Retailers have legitimate interests in preventing crime, but the routine collection of biometric information from the public is a serious step. The fact that facial recognition can be installed does not automatically make its use necessary or proportionate.
The debate is ultimately about the kind of public environment people want. If biometric identification becomes an ordinary condition of entering shops, shopping centres and other everyday spaces, the practical ability to move anonymously may shrink. The challenge for regulators and businesses is to protect workers and customers without treating every visitor as a potential suspect.
Sources and further reading
- Biometrics and Forensics Ethics Group: Public-private collaboration in live facial recognition (May 2026)
- Information Commissioner’s Office: Facial recognition technology and surveillance
- ABC News: How facial recognition could differ from CCTV in supermarkets (27 August 2026)
- The Guardian: Privacy concerns over supermarket facial recognition trials (18 August 2026)
- The Guardian: Facial recognition cameras in British retail centres (29 July 2023)
