What is live facial recognition (LFR)?

Live facial recognition is a surveillance technology that scans faces captured by a live camera feed in real time, and automatically compares each one against a pre-compiled list of people the police are looking for — the "watchlist." When the system judges a face to be a likely match, it generates an alert; an officer then reviews the alert and decides whether to stop the person.

This page is a plain-language, factual overview. It doesn't offer legal advice, and it doesn't try to settle the policy debate around this technology — it exists so readers can understand what the data on this site is actually describing.

How a deployment works

1. Watchlist creation. A police force compiles a list of individuals it's looking for. Police say watchlists include people wanted on warrants, individuals subject to court orders and people reported missing, but their exact composition is rarely disclosed, so it's often unclear precisely who is on them, or why. Watchlist size also varies by deployment and force.
2. Camera deployment. Cameras are set up in a public location — typically mounted on a police vehicle or a temporary rig — for a fixed window of time, with signs nearby informing the public that LFR is in operation.
3. Live matching. As people pass the cameras, the system extracts a biometric template from each face and compares it against the watchlist in real time.
4. Alert and review. A likely match triggers an alert to an officer, who reviews the underlying images before deciding whether to approach the person. The system does not act automatically — a human makes the final call.

Key terms used on this site

• Deployment — one use of LFR at a specific place and time; the basic unit each row of data on this site represents.
• Watchlist size — the number of people on the watchlist for that deployment, where a force discloses it.
• Faces scanned — the number of faces the system processed and compared against the watchlist during a deployment. Not every force records or discloses this per deployment — see the Methodology page for how gaps and estimates in this figure are handled.
• Alert — a system-generated potential match, flagged for human review.
• True / false alert — whether an alert is later judged, by an officer, to be a correct match or not. Definitions and review processes vary by force.

Context worth knowing

LFR use by UK police has expanded significantly since its early trials, and remains a live policy debate. Points raised on different sides include:

• Police forces and the Home Office have generally framed LFR as a tool for locating wanted individuals and preventing crime more efficiently than manual identification.
• Civil liberties groups and some independent researchers have raised concerns about mass biometric surveillance of the public, accuracy differences across demographic groups, and the adequacy of the legal framework governing its use.
• The technology's legal basis has been tested in UK courts — most notably in Bridges v South Wales Police (2020), in which the Court of Appeal found the force's use of LFR unlawful. Its rules left too much discretion to individual officers over who could be put on a watchlist and where the cameras could be deployed. The court also found the force's data protection impact assessment was deficient, and that it hadn't done enough to check whether the software was biased on grounds of race or sex. The ruling didn't rule out LFR in principle, and forces have since adjusted their policies in response.

For how the underlying figures are collected and harmonised across forces, see Methodology.