British Transport Police facial recognition trial produced one false alert and no alert-led arrests

British Transport Police's six-month live facial recognition trial in London railway stations scanned more than half a million faces but produced one incorrect watchlist alert and no arrests resulting from an alert, according to an FOI document reported by the Guardian on Tuesday. The result raises a practical question for a pilot that has since expanded: what benefit does the technology deliver in busy stations relative to its cost and the number of people scanned?

The document, obtained by Liberty Investigates and shared with the Guardian, put equipment hire and police staffing for 18 deployments between February and July at £320,786. The deployments also used almost 100 hours of police officers' time. Those figures describe the reported trial deployments; they do not establish the cost of the wider programme or of any later expansion.

What the first six months showed

Live facial recognition compares camera images against a police watchlist and alerts officers to possible matches. BTP launched the railway pilot to explore whether the technology could help identify wanted offenders and people who might pose a risk to passengers and staff. Its first six-month phase generated one alert against the watchlist, which proved to be a false match, the Guardian reported. The reported result is an account of that phase's watchlist alerts, not a complete measure of all police activity at the stations.

BTP told the Guardian that officers made arrests during deployments for offences including assault, theft, possession of an offensive weapon and public-order offences. Officers also located people wanted by courts or other forces, it said. The force stressed that those actions did not result directly from facial recognition alerts, so it excluded them from the technology's performance data. That distinction matters when judging what the cameras themselves achieved: police activity at a deployment site is not evidence that a watchlist match produced an arrest.

An expanded pilot and different results

BTP extended the trial for four more months and expanded deployments to include London Underground stations, the Guardian reported. Since that extension, the force said, three alerts have been confirmed as matches. The people involved were subsequently confirmed to be complying with sexual harm prevention orders or other court conditions. These were later alerts, separate from the one incorrect alert in the first six months; the account does not establish arrests or other enforcement action resulting from them.

Transport for London backed the extension, saying watchlists at selected stations could help tackle violence against women and girls. That is a stated aim of the deployment, rather than a demonstrated outcome. BTP has said it is testing how the technology can support identification of people of concern on the railway network. A full evaluation of the expanded pilot would need to address its results as well as its intended purpose.

Cost, privacy and oversight

Fraser Sampson, a former UK biometrics and surveillance camera commissioner, told the Guardian that judging proportionality depends on where and when cameras are deployed, who is on the watchlist and whether those people are likely to pass the cameras. He said the trial did not appear fruitful if success is measured by people caught through police alerts. The Guardian identified Sampson as a non-executive director of Facewatch, a company that operates facial recognition systems in shops; his assessment is not a completed official evaluation of BTP's pilot.

Labour MP Bell Ribeiro-Addy called live facial recognition a serious invasion of privacy and urged a pause in its rollout until stronger safeguards, including a legal framework, are in place. Her objection concerns the wider use of the technology as well as this trial's outcome. The debate extends beyond whether a camera can find a match: it also concerns the justification for scanning large numbers of travellers in pursuit of people on a watchlist.

In an August review of facial recognition in policing, the Information Commissioner's Office said false matches can lead to wrongful intervention, accusation or arrest. Its audits of five police forces in England and Wales found inconsistent data-protection compliance and prompted 107 recommendations, all accepted or partially accepted by the audited forces. The ICO identified a need for stronger oversight, record keeping, accuracy checks and steps to reduce unfairness or bias. Those audits provide context for the risks and safeguards at issue; they were not findings about BTP's London pilot.