London’s Metropolitan Police Service (MPS) says the hundreds of live facial recognition (LFR) deployments across the Capital last year led to 962 arrests, according to a new report on the controversial tech’s use.
The report [PDF] covered the period from September 2024 to September 2025, noting that from the 203 deployments, LFR cameras triggered 2,077 alerts with 10 false positives.
Cops said the arrests were mostly comprised of people who were wanted by the courts (549); and those the MPS said it had “reasonable grounds” to believe may be either about to commit, currently committing, or have already committed an offence (347). The remaining 85 arrests related to those managed by multiple agencies (MAPAA nominals), such as registered sex offenders (RSOs), and stalkers, as well as those in breach of their conditions.
It said more than a quarter related to individuals involved with violence against women and girls, and that none of the total arrests were officially deemed to be unnecessary.
Lindsey Chiswick, lead for LFR at the Met and nationally, said: “We are proud of the results achieved with LFR. Our goal has always been to keep Londoners safe and improve the trust of our communities. Using this technology is helping us do exactly that.
“This is a powerful and game-changing tool, which is helping us to remove dangerous offenders from our streets and deliver justice for victims.
“We remain committed to being transparent and engaging with communities about our use of LFR, to demonstrate we are using it fairly and without bias.”
About those false positives
The MPS chose the most attractive way of interpreting the false alert rate, saying that it represents just a 0.0003 percent failure rate, considering the total 3,147,436 faces it scanned across all deployments.
But, if you look at it in terms of the number of alerts LFR cameras made (2,077), it represents a considerably less appealing rate of 0.48 percent.
The MPS assured that only six of these 10 false positive alerts led to engagements with the subjects, each lasting no more than five minutes, and said none of the people were arrested.
Ten different reasons were given for each of the false positives, which mainly related to image quality issues such as poor lighting, poor angles, and faces being covered by clothing or other passersby.
However, in one case, an LFR camera falsely alerted police to an identical twin of a genuine suspect, and another got the gender of the subject wrong, the report says.
One aspect of the report that caught the attention of privacy and human rights groups was the ethnic breakdown of these false positives, with eight out of ten concerning Black people.
The report states:
LFR and ethnic biases
Ethnic biases have been central to the longstanding criticisms of LFR tech globally, and the UK’s previous use was no stranger to similar scrutiny.
On multiple occasions, according to separate research in 2020 and 2023, the UK’s LFR technology was shown to exhibit racial biases in false positive alerts.
Regardless, the MPS has consistently defended the technology’s performance, despite being aware of the issues.
The latest annual review of LFR deployments goes on to claim the 80 percent Black rate for false positives was nothing to worry about.
“Overall, the system’s performance remains in line with expectations, and any demographic imbalances observed are not statistically significant. This will remain under careful review.”
Demographic biases in LFR alerts are independently tested at the National Physical Laboratory at 0.60, 0.62, and 0.64 match thresholds.
The threshold for the reporting period was set between 0.60 and 0.64, meaning that a positive alert would only be made if a subject was face-matched within that threshold, and the ten false positives were all made when that was set at 0.64.
The report goes on to say the overrepresentation of Black men in the data (seven of the eight false positives were Black males) could be explained by the location of LFR deployments.
“Whilst the difference is not statistically significant, one factor that may influence LFR outcomes more generally is the location of deployments, which are focused on crime hotspots,” it said.
“These areas often overlap with communities experiencing higher levels of deprivation, where Black males are statistically overrepresented in both crime and victimization data.”
Long-term opponents to LFR use in the UK, Big Brother Watch’s legal and policy officer Jasleen Chaggar scathed the report’s findings.
“It’s disturbing that 80 percent of the innocent people wrongly flagged by facial recognition were Black,” she said. “We all want police to have the tools they need to cut crime but this is an Orwellian and authoritarian technology that treats millions of innocent people like suspects and risks serious injustice.
“No law in this country has ever been passed to govern live facial recognition and given the breathtaking risk to the public’s privacy, it is long overdue that the government stops its use to account for its serious risks.”
Broad support
The MPS said a Public Attitude Survey, commissioned by the Mayor’s Office for Policing and Crime, revealed that 85 percent of Londoners support the use of LFR across the city, although that comes with some caveats.
The LGBT+ community was the strongest opponent compared to the average, followed by those of mixed ethnicities, then by Black respondents.
The survey showed younger people were more likely to oppose the technology than older generations, with the strongest deviation from the average coming from those aged 25 to 34. The over-65s were the most supportive age group by a distance.
The MPS said it understands the importance of building trust in new technologies and said it was committed to working with the wider community on that front.
The perceived success of LFR has led the government to encourage its wider use more broadly across the UK.
Croydon’s use of the technology, which is unlike many other temporary deployments in that it has two permanently installed cameras, will inform new literature set to be released this year to guide other regions’ LFR strategies. ®