Live facial recognition (LFR) scans faces in a real-time camera feed, usually mounted on a van or fixed pole at a public location, and compares each face against a watchlist of people police are looking for. A match generates an alert for an officer to review before any action is taken; a non-match is supposed to be deleted immediately without being stored.

The core legal question in the UK, settled in outline by the Court of Appeal's 2020 Bridges ruling but still argued over in detail, is whether the watchlist compilation, the deployment site selection, and the handling of the many people who are scanned but never matched are all sufficiently governed by clear, foreseeable rules. Forces have revised their policies since Bridges, and campaigners maintain the revisions still leave too much discretion to individual officers and forces.

Accuracy varies significantly by system, lighting, camera angle, and the demographic composition of the watchlist and the crowd being scanned. Independent testing by the National Physical Laboratory found the Metropolitan Police's system, when properly configured, produced very low false-positive rates at the settings the force actually uses, though critics note real-world deployment conditions are harder to control than a laboratory test.