Leadership1 publisher3 min readPublished
British Transport Police extends a £320,786 face-scanning trial that yielded one wrong alert
British Transport Police is extending a £320,786 facial recognition trial that scanned over 500,000 faces and produced one false alert and no arrests. Its public defence leans on arrests officers made during deployments that its own performance data leaves out.
The Board Room · Leadership desk
What happened
- The spending covered equipment hire and police staffing for 18 deployments between February and July, according to a freedom of information document obtained by Liberty Investigates.
- The trial also absorbed almost 100 hours of police officers' time across London's railway stations.
- BTP announced last month that it would run the trial for four more months and extend deployments to London Underground stations.
- More than half of police forces in England and Wales have now deployed live facial recognition, according to an analysis by Liberty Investigates.
Compiled by The Board RoomSomething wrong?How this is made
Why it matters
- decision BTP counts only arrests that follow an alert. For the extension to show a result, the cameras have to produce correct matches, or the force has to argue from arrests it chose to exclude.
- exposure Oversight boards judging proportionality will weigh the processing of more than half a million faces as special category data against a single incorrect match.
- contradiction TfL backed the extension saying stations are "chosen for maximum impact", yet in the first six months some of the busiest hubs in London yielded one match, and it was wrong.
The board-deck version of this trial is a unit cost. Spread across 18 deployments, £320,786 comes to about £17,800 each [1]. Spread across more than half a million faces, it is under 64p a face [2]. Divided the same way, output is fewer than one alert per 500,000 faces, and the rate of correct matches is zero [5].
BTP's public defence leans on arrests the cameras did not produce. A spokesperson said officers made "a number of associated arrests" during deployments, for offences including assault, theft and possession of an offensive weapon [15]. "As these arrests did not result directly from an LFR alert, they are not included within LFR performance data," the spokesperson said [16]. The rule is sound. It separates what the camera found from what the officers standing next to it found, and by that count the camera found nothing that led to an arrest [1].
The sharpest criticism came from inside the industry. Fraser Sampson, the former biometrics and surveillance camera commissioner for the UK, is now a non-executive director of Facewatch, which runs facial recognition in shops [6]. "Success for the police trying to catch people means people being caught. In that respect the trial doesn't appear to have been very fruitful," he said [7].
BTP's answer is that a pilot exists to learn. The force said it launched the trial "to understand how the technology could most effectively support the identification of wanted offenders and individuals who may pose a risk to passengers and staff" [14]. It also said it had "continued to refine how the capability is deployed within the railway environment" [17]. Learning is a fair basis for an extension if the lessons are written down. The Guardian's account does not say whether BTP set a success threshold before the first deployment, or what the refinements changed.
Sampson also named the variable that decides results. "Everything depends on the police watchlists; if you're not on a watchlist, the live FRT cameras do not 'see' you and they don't retain your image in the same way as CCTV cameras," he said [8]. Adding Underground stations raises the number of faces passing the cameras [4]. A correct alert still needs someone on the list to walk past one.
This quarter's decision is four more months. If they run at the first phase's average of about £53,500 a month, they will cost roughly £214,000 [3][4]. Next quarter's consequence is a programme that has run for ten months [6], and I'd expect it to be reviewed then as a standing service, with the first six months as its baseline. It will sit in a city where the Met and the mayor's office plan fixed cameras in the West End, and Sadiq Khan has confirmed they will be used on Oxford Street [11].
Opposition in Parliament is on the record. The joint committee on human rights named the rollout of live facial recognition a "particularly clear example of risk" [12]. Bell Ribeiro-Addy, Labour MP for Clapham and Brixton Hill, said: "Live facial recognition technologies are a serious invasion of our privacy. The government should be suspending their rollout until there are stronger safeguards in place, including a proper legal framework." [13]
What to watch
- Whether BTP publishes alert, match and arrest figures for the four-month extension, including the new Underground stations.
- A correct alert that leads to an arrest during the extension would be BTP's first result under its own counting rule.
- Whether the government brings in the law the joint committee on human rights has called for before the Met's fixed Oxford Street cameras go live.