1.
Imagine rushing through a crowded London railway station, your train already announced on the board, your eyes fixed on the departure screen. You pass a cluster of cameras mounted high on a wall and give them no thought—they look like ordinary CCTV. But these are not ordinary cameras. They are part of a live facial recognition trial, scanning your face and instantly comparing it to a watchlist of people police want to find. For months, from February to July this year, British Transport Police ran this experiment in some of the capital’s busiest transport hubs. Their stated goal was noble: catch offenders, protect passengers, and disrupt serious crime before it happens. But the reality, as revealed by a freedom of information document obtained by Liberty Investigates and shared with the Guardian, tells a much more complicated and unsettling story. The trial cost more than £320,000, consumed almost one hundred hours of police officers’ time, and scanned over half a million faces. In return, it produced exactly one alert against a police watchlist—and that alert was a false positive. No one was arrested as a direct result of the technology. No wanted fugitive was plucked from the crowd. No court order was enforced in that time. Instead, the trial quietly demonstrated that high-tech surveillance can be expensive, time-consuming, and still leave ordinary commuters feeling like suspects in their own daily journeys.
2.
The details of the trial, buried in an innocuous-looking spreadsheet and later confirmed through the official information request, are striking. There were eighteen separate deployments across London’s railway network, each requiring specialist equipment hire and teams of police officers to monitor the feeds and respond to any potential matches. The total price tag for equipment and police staffing was £320,786. During those eighteen deployments, the cameras scanned more than half a million faces. If the technology had worked as advertised, this would have been celebrated as a breakthrough in proactive policing. Instead, the only match to a watchlist was a false positive—an incorrect identification that sent officers running after someone who had done nothing wrong. The system was aimed at catching people who were wanted for crimes or breaching court conditions, but it found nobody. That is the core problem with live facial recognition in public spaces: it turns every passerby into a potential data point, and yet its ability to actually help police is still unproven. The trial was supposed to be a rigorous test, but to many, the result looks less like evidence of success and more like a warning sign. It is particularly troubling when you consider that the technology processes what UK law calls “special category personal data”—sensitive biometric information that is considered one of the most intimate forms of personal data. For the vast majority of people scanned, this happened without their knowledge or consent, simply because they were going about their daily lives.
3.
What makes the trial even more puzzling is the decision to extend it. Last month, British Transport Police announced that the live facial recognition pilot would continue for another four months and expand to include London Underground stations. After the extension, police said there had been three positive alerts, all involving people who were subsequently confirmed to be complying with sexual harm prevention orders or other court-imposed conditions. That sounds like a small success, but it raises an uncomfortable question: did police really need to scan hundreds of thousands of faces in order to discover that three people had been behaving themselves? The stated purpose of the expanded trial, according to Transport for London, was to target and identify people on police watchlists at key stations chosen for “maximum impact,” with a particular focus on tackling violence against women and girls. That is an important and urgent goal, as anyone who has ever felt unsafe on public transport can understand. But the gap between the promise and the delivery is enormous. Fraser Sampson, a former UK biometrics and surveillance camera commissioner, now a non-executive director of Facewatch, put it plainly: police need to prove the technology is proportionate. Success in a shop, he said, means shoplifters stay away. Success for police trying to catch people on trains and platforms means people are actually caught. By that standard, the trial “doesn’t appear to have been very fruitful.” He also reminded the public that the cameras only “see” people on a watchlist, and that the images of everyone else are not retained in the same way as CCTV footage. But that assurance does not erase the discomfort many feel about being silently analysed by a biometric system.
4.
The London railway trial is not an isolated experiment. According to an analysis by Liberty Investigates, more than half of police forces in England and Wales have now deployed live facial recognition somewhere on British streets. The Metropolitan Police and the London mayor’s office have announced plans to install fixed live facial recognition cameras in the capital’s West End, and Mayor Sadiq Khan has confirmed the technology will be used on the newly pedestrianised Oxford Street. This is not some distant future; it is happening now, in one of the most visited shopping streets in Britain. The expansion has alarmed civil liberties campaigners and even parliamentarians. A recent report from the parliament’s joint committee on human rights identified the rollout of live facial recognition as a “particularly clear example of risk,” and called on the government to introduce a law to protect human rights in the age of artificial intelligence. The concerns are not simply ideological. They are practical: Is the technology accurate enough? Are the watchlists fair and free from bias? Who determines who gets placed on them? What happens when someone is falsely identified—do they even know? In a country that prides itself on freedom and privacy, the idea of being tracked by a camera that can recognise you in a crowd is deeply unsettling. Bell Ribeiro-Addy, the Labour MP for Clapham and Brixton Hill, said live facial recognition technologies are “a serious invasion of our privacy” and called on the government to suspend their rollout until stronger safeguards, including a proper legal framework, are in place.
5.
Unsurprisingly, British Transport Police defend the trial. A spokesperson said the pilot was launched “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.” They also pointed out that, during the deployments, officers made a number of associated arrests—including for assault, theft, possession of an offensive weapon, breach of a criminal behaviour order, and public order offences—and that they located individuals wanted by the courts and other police forces. But, importantly, they acknowledged that these arrests did not result directly from a live facial recognition alert, and so they are not included in the LFR performance data. That is a crucial distinction. If officers at a station are making arrests anyway because they are physically present, then the cameras are not doing the work. They are simply an expensive backdrop to ordinary police activity. The spokesperson also explained that the watchlist growth has been “deliberate, controlled and aligned to BTP policy,” and that they began with a limited watchlist while ensuring data quality, governance, and safeguarding were robust. That sounds reassuring, but it also reveals a truth: the system is not simply about catching evildoers. It is a new method of social control, one that requires constant refinement and testing precisely because it is so powerful and so easy to get wrong. Every step forward is being taken without clear legislative consent, without comprehensive public debate, and without a settled answer to the question of whether the benefits justify the intrusion.
6.
So what is the takeaway from this costly, largely ineffective trial? It is not that police should never use technology; there are cases where facial recognition is already used in a much more focused and successful way, such as identifying people from CCTV footage after a serious crime. But live facial recognition in public places is a different animal. It is proactive surveillance of everyone, not just suspects. It turns every commuter, every tourist, every person simply trying to get home into a potential data point. The human impact is profound. A woman travelling alone at night might welcome the idea that police are watching, but she might also feel uneasy knowing that the system is not watching out for her—it is watching her, comparing her identity to a list of people suspected of crimes she has never committed. The false positive in the London trial was not just a technical glitch; it was a reminder that behind every error is a real person who could be stopped, questioned, and publicly embarrassed for no reason. If the technology is to have any future, it must be accompanied by a proper legal framework, independent oversight, and transparent reporting. It must be proportionate, necessary, and genuinely effective. For now, the evidence from London’s railway stations is disappointing. More than half a million faces scanned. Hundreds of thousands of pounds spent. Nearly one hundred hours of officers’ time dedicated to monitoring. And at the end of it all, one false match and no arrests. That is not the kind of result that should be rewarded with an expansion. It is the kind of result that should make politicians pause, ask searching questions, and remember that the right to move through the city without being digitally examined is a freedom worth protecting.