London Underground live facial recognition runs against a watchlist of wanted persons

British Transport Police are running live facial recognition against a watchlist of robbery, knife-crime, and sexual-assault suspects at selected London Underground stations. The system runs under a BTP policy published in February 2026. The wider national biometrics law is still being consulted on by the Home Office.

BTP's cameras do not record everyone who passes. They compare each face against a curated list of people BTP wants to arrest. A match generates an alert and an officer intercepts the person. A non-match is deleted "automatically and immediately" as the passenger continues their journey, according to the joint BTP/TfL trial notice. The watchlist is scoped to serious, high-harm offences: robbery, knife crime, and sexual offences.

In a TfL survey, 77% of 13-to-16-year-old girls (116 of 150) said they had experienced unwanted sexual behaviour on a public transport journey in the previous 12 months; 91% (136 of 150) said they had witnessed it; and 40% (400 of 1,006) said they had chosen not to make a journey because they felt unsafe, according to the BTP/TfL trial notice. Women and girls in the same survey named using LFR to identify offenders as a priority for action. The trial is part of TfL's wider violence-against-women-and-girls safety programme.

The policy text behind the cameras is BTP's Overt Live Facial Recognition Policy v1, published 10 February 2026. It is the first concrete, public BTP document setting out the legal basis, watchlist criteria, signage obligations, and data-handling rules for the technology. The policy describes the watchlist in category terms, as wanted persons linked to serious offences, but it does not enumerate which stations will host the cameras, how many cameras each station will run, or how individual entries are added to or removed from the list. Station selection is described as intelligence-led, drawing on crime data and operational analysis, but the criteria are not in the public document.

Live facial recognition (LFR) is real-time face-matching on live video, in public, against passers-by who have not been individually suspected of anything. It is distinct from retrospective face search, in which investigators run still images against police databases after the fact. Independent testing commissioned by the Metropolitan Police on earlier-generation LFR has found higher false-positive rates for Black faces than white faces, and the UK Parliament's POST briefing on facial recognition in policing sets out the ongoing parliamentary concerns around demographic accuracy, scope creep, and the legal authority for live face-matching in public spaces. The BTP/TfL trial notice does not publish demographic error rates for the current deployment.

The wider statutory ground is still being set. The Home Office consultation on a new legal framework for law enforcement use of biometrics, facial recognition, and similar technologies is the venue where consistent national rules on accuracy thresholds, watchlist governance, retention, and oversight would, in theory, be settled for every UK force that wants to run LFR. Until that framework lands, BTP's deployment sits on BTP's own published policy plus the existing common-law and statutory powers of constables.

Oversight runs through BTP's policy plus the existing policing complaints route. The policy sets out signage, an obligation to brief officers before deployment, and a complaint pathway, but it does not create an independent review body for the LFR programme specifically. The POST note sets out what parliamentary researchers describe as best practice for any UK force running overt LFR.

Until the Home Office consultation lands, the rulebook for live face-matching on the Tube is the policy of the force running it.