In London, cameras recognized over half a million faces. Result: one "criminal". And not even the right one

The pilot project of the British Transport Police launched in February 2026. Cameras were installed at major London stations, comparing passengers' faces in real-time with police lists. This way, the police aimed to identify people who were wanted or those on whom the court had imposed certain restrictions. If the program finds a match, it signals a police officer, but the final decision must still be made by a human. Images of people not on the list, according to the police, should be immediately deleted.

In practice, the first six months yielded almost zero results. Documents obtained by Liberty Investigates from the police via a freedom of information request show that the system was launched 18 times. During this period, it scanned over 500,000 faces, and the rental of equipment and staff work cost approximately $426,000. Additionally, the police spent almost 100 working hours on these operations.

During the entire period, the algorithm only once reported that it had allegedly found a person from the watchlist. However, after manual verification, it turned out to be a different person.

Police decided to continue anyway

Despite this outcome, the Transport Police did not stop the experiment. In August, it was extended for another four months and expanded to London Underground stations. The police explain that the first phase was necessary not only to apprehend people but also to understand how best to use such technology in the transport system. During the test, camera locations, operating procedures, equipment, and the lists of people the system should search for were changed.

After the project expansion, three confirmed matches appeared. In all cases, the system correctly identified people who were subject to court restrictions. However, checks showed that they were not violating these conditions, so there were no grounds for their arrest.

At the same time, the police emphasize that patrols near the cameras were not entirely fruitless. Officers arrested people for assault, theft, carrying prohibited weapons, and public disorder offenses. However, these arrests occurred not thanks to the facial recognition system, and therefore were not included in its statistics.

Is there any point to such a system?

Former British Biometrics and Surveillance Camera Commissioner Fraser Sampson pointed out that the technical ability of a system to recognize faces alone does not mean that its use will be effective. Much depends on where the cameras are placed, when they are switched on, who is on the watchlist, and how likely it is that these people will actually be in a specific location.

Sampson compared the police use of such technology to cameras in retail establishments. For a shop, a good result might already be that a known thief was afraid to enter. But if the police install cameras to find specific people, the result is logically measured by the number of those who were actually identified. In this sense, he noted, the first six months of the London trial can hardly be called fruitful.

However, the debate around the technology is not only about its effectiveness. The system processes biometric data of all people who fall within the camera's field in real-time, so human rights defenders demand that the police justify why such surveillance is necessary and whether its scale corresponds to the actual results.

The issue is becoming increasingly relevant as similar systems spread. According to an analysis by Liberty Investigates, facial recognition on the streets has already been used by more than half of police forces in England and Wales. In London, stationary cameras are also planned for the city center, including Oxford Street, which was recently pedestrianized.

• 1
• 1
• 1
• 0
• 0
• 0
• 0
• 0