London: A six-month trial of live facial recognition (LFR) technology at London railway stations cost more than £320,000, scanned more than 500,000 faces and produced only one alert against a police watchlist, which turned out to be a false match.
The British Transport Police (BTP) conducted the trial between February and July across 18 deployments at some of the capital’s busiest transport hubs. The technology was tested as a way of identifying wanted offenders and people breaching court orders.
Freedom of information data obtained by Liberty Investigates and shared with the Guardian showed that equipment hire and police staffing for the deployments cost taxpayers £320,786. The single watchlist alert did not lead to an arrest because it was an incorrect identification, known as a false positive.
BTP announced last month that it would extend the trial by another four months, while also expanding deployments to London Underground stations.
When Transport for London (TfL) announced its support for the extension, it emphasized that LFR would be used at selected key stations to identify people on police watchlists. TfL said the technology would also specifically target violence against women and girls, whose travel patterns can be affected by experiences of sexual harassment and sexual offences.
Fraser Sampson, a former Biometrics and Surveillance Camera Commissioner, remarked that police needed to demonstrate that LFR deployments were proportionate. Sampson added that several factors could affect the effectiveness of the technology, including the location and time of deployment, the composition of the watchlist and the likelihood of people on the list being present.
The Commissioner, who is now a non-executive director at facial recognition company Facewatch, noted that the use of LFR in public transport environments was different from its deployment in shops. While preventing a suspected offender from entering a shop could be considered a successful outcome, police deployments are intended to identify and apprehend people.
The effectiveness of railway deployments therefore depended heavily on the police watchlists. People who are not included on a watchlist are not identified by LFR cameras, and their images are not retained in the same way as images captured through conventional CCTV.
Sampson also emphasized that LFR processes special-category personal data, meaning its use must be appropriate, proportionate and necessary. He added that police must demonstrate that deployments meet those requirements and that oversight boards should scrutinise how the technology is being used.
The Metropolitan Police and the London mayor’s office announced this summer that fixed LFR cameras would be installed in the capital’s West End. London Mayor Sadiq Khan has also confirmed that the technology will be used on the newly pedestrianised Oxford Street.
A recent report by Parliament’s Joint Committee on Human Rights, which called for legislation to address the risks artificial intelligence poses to human rights, identified the expansion of live facial recognition as a particularly clear example of such risks.
Labour MP Bell Ribeiro-Addy described LFR technology as a serious privacy concern and called for its rollout to be suspended until stronger safeguards, including a proper legal framework, are established. BTP stated that it launched the pilot to assess how LFR could best support identifying wanted offenders and people who may pose a risk to passengers and staff.
The force noted that officers made several arrests during the deployments for offences including assault, theft, possession of an offensive weapon, breaches of criminal behaviour orders, and public order offences. Officers also located individuals wanted by courts and other police forces.
However, BTP clarified that these arrests and other outcomes did not result directly from LFR alerts and therefore were not included in the LFR performance data.
Throughout the pilot, BTP stressed that it refined how the technology was used in railway environments, including changes to deployment locations, operating procedures, equipment and watchlist construction.
The force also noted that the growth of its watchlists had been deliberate and controlled, with the initial list kept more limited while BTP worked to ensure data quality, governance and safeguarding arrangements were robust.
MOST READ | Is double-dipping unhygienic? Experts explain the risks
