Retailers question using live facial recognition for shoplifting

Retailers have told the House of Lords that live facial recognition (LFR) would be of limited use in tackling shoplifting due to the associated safety and ethical concerns, but that working with police to automatically identify offenders after the fact with retrospective versions of the technology should be standard practice.

In May 2024, the House of Lords Justice and Home Affairs Committee (JHAC) launched an inquiry into combatting shoplifting, which partly focuses on how police and retailers are using both live and retrospective facial recognition (RFR) to deal with retail crime.

Speaking about the Co-op Group’s approach to facial recognition in stores on 3 September, Paul Garrard, the organisation’s public affairs and board secretariat director, told Lords that while Co-op Group itself does not use LFR to detect shoplifting in real-time, it will compile an “evidence pack” for police when reporting a theft, which will include material like CCTV and staff body-worn camera footage to be run through RFR software.

He added that although some police forces will take the compiled footage and compare it to photos contained in the Police National Database (PND) – which holds millions of custody images, many of which are being unlawfully retained by the Home Office – it is not currently common practice for police to automatically check the images provided against the database.

“I think it should be – we’ve seen in South Wales that it really helps,” he said, adding that when police do run the images through the PND, they will often realise that they are also “interested in them for other reasons”.

In October 2023, the UK government launched a business-police partnership called Project Pegasus, part of which revolves around 14 of the UK’s biggest retailers – including M&S, Boots and Co-op – sharing CCTV footage with forces so they can run it through the PND using RFR software.

Speaking during the shoplifting inquiry’s previous session, chief constable Amanda Blakeman told Lords that RFR searches are a major tool in dealing with shoplifting, noting that it is especially helpful in identifying people who are unknown to police but who have still been through the custody process.

“If somebody is known, there is no need to put them through the retrospective facial recognition; we already know who they are,” she said. “From that perspective, we are working hard with forces to make sure that those reasonable lines of inquiry include exploiting opportunities to look at the evidence that we already have retained on police systems.”

She added: “Facial recognition forms an ability to prevent and an ability to gather intelligence. It provides us with an important opportunity to identify some of those prolific offenders. It has certainly been an important part of the work … in relation to those individuals who travel up and down the country and appear at perhaps eight or nine different locations in forces, and of our being able to identify who they are and get some good intelligence around them.”

However, despite highlighting the benefits of RFR to retailers, Gerrard noted during the latest committee session the Co-op group has “no plans” to introduce LFR in stores, as “we can’t really see what intervention it will drive [that would be] helpful”.

This is partly on the basis that it would place the onus on store staff to intervene if a shoplifter was flagged by an LFR system, therefore placing them at greater risk of violence or abuse, but also partly because locally created facial-recognition watchlists would be largely ineffective compared to the PND.

Even with the PND, Gerrard said that there are many reasons people could be included in the database, many of which have nothing to do with shoplifting offences, adding: “There are ethical issues we’d want to think through.”