This week we monitored a met police live facial recognition deployment in Croydon. 83% were wrongful stopped, and we witnessed systemic racism with most black suspects (all innocent) being immediately handcuffed before checks were carried out despite fully cooperating with the police – one shackled for over 30mins, whilst white suspects were never handcuffed before checks were carried out. Black innocent suspects were on average stopped for over twice as long as white innocent suspects.
Most Surveilled In UK
The Met Police have deployed facial recognition surveillance in Croydon over 20 times this year, more than any other place in the whole of the UK. On Wed 28th August 2024 they once again targeted our community for surveillance, rolling out their live facial recognition van into Croydon town centre. This was despite over 20 community organisations and local councillors writing the the Met urging them to stop facial recognition surveillance of our community. This deployment followed a three month lull with no deployments in our borough since 29th May. We attended the deployment to observer and monitor what happens.
To comply with the Mets operating procedures, instituted after the 2020 legal ruling which found that the Police’s use of intrusive and discriminatory facial recognition violates privacy rights and breaks data protection and equality laws, they are meant to give advanced notice on social media. For that to be meaningful it would be reasonable to give notice the day before, but they chose instead to give only 30 minutes notice – a tweet sent at 11:30 for a deployment starting 12 noon. By the time we realised what was going on and got there it was 2:30pm. The deployment finished around 6pm. This is an account of what we saw in those 3.5 hours.
Deployment
The deployment used one van. Apparently the Met at the moment has only 2 LFR vans, the other was being used in Westminster at the same time. The van was positioned outside Barclays Bank, hidden at the intersection of North End, Church Street, High Street and George Street. The van was armed with six sets of stereo vision cameras – 3 pointing down Church Street (surveilling people coming from the busy market), 2 aimed at George Street (targeting people coming from East Croydon station) and 1 pointing at the High Street (surveilling shoppers).
Each of these streets had “Facial Recognition in use notices” as stipulated by the operating procedures laid out by the College of Policing, though with everything else on display you wouldn’t readily see these small A3 signs tied to a lamppost, unless you were looking for them like we were.
We noticed two of the notices had suspicious people hanging around nearby, a woman near the High Street notice pretending to be fiddling on her mobile was particularly obvious. Our suspicions were confirmed as later both were revealed to be plain cloths police officers. Whilst the notices say “there is no legal requirement for you to walk through the LFR system” the reality is that if you don’t then that’s construed as suspicious behaviour, and these plain cloths officers are there to stop you and force you to submit to a scan. It doesn’t seem to occur to the Met that ordinary people, not wanted for any crime, may not want to walk through a police biometric checkpoint, have their faces scanned and biometrics extracted without their permission.
Reaction
Initially when I arrived I could feel their eyes were on me, it was only after I took out the “stop facial recognition” poster that they switched from all eyes on me to ignoring me.
The poster did elicit some great responses from the public. Some people thanked us for being there; others recognising Mark Rowley, Commissioner of the Met police, on the poster being scanned saw the humour and laughed with a thumbs up; and some concerned about the erosion of their rights engaged in conversation. One person lamented “they are taking everything from us, and there is nothing we can do”. Desperation and a sense of helplessness was a common theme, a woman leaving the market seeing me with the poster near the facial recognition sign said “it’s too late, it’s everywhere now, we can’t stop it, we should have stopped it years ago”. I replied “we have to start somewhere, today the cameras are on a van – a temporary installation, if we dont stop them they will have them permanently pointing at us”.
Stops
During the 3.5 hours we witnessed 6 people being stopped by the police, with only one being arrested. The other 5 ( 83% ) were wrongly stopped (and often searched and handcuffed), harassed and falsely accused of something they were not wanted for. There was a 7th alert but the police couldn’t locate the person (despite 4 officers scouring George Street).
Accuracy of facial recognition algorithms and the accuracy of police data used to compile watchlists are of concern as they have both been shown to be deeply flawed. Often the algorithm identifies the wrong person who is not on the watchlist (this is especially true for people of colour). Sometimes the match is so bad that it is rejected by officers and no stop will be made, and we will not hear of the erroneous match. Though this is rare as officers seem to trust the AI more than their own common sense – in London Bridge when the algorithm matched a black man to a criminal suspect, despite missing a scare on his nose and missing tattoos, both of which the officers noticed, and despite him producing a copy of his passport to confirm his identify, they still held him for half an hour until the system cleared him! This is hardly surprising when senior officers extoll the virtues of LFR as being “a perfect system” (Ealing Met lead on LFR, July 2024). Other times a person has been wrongly put on the watchlist – for example claiming there is a warrant for their arrest when in reality there is no warrant – they are not wanted for anything.
Interviewing those falsely stopped revealed that it was flawed data more than errors in the algorithm that caused the false stops on this day.
We believe the algorithm was running at the highest threshold of 0.64. The Met police was previously using threshold of 0.6, claiming that there was no statistical bias at 0.6 and the error rate was just 1 in 6,000. These have proved to be false, both by independent academic studies and in practice revealed by the Mets own field results (85% of the time it gets it wrong, based on data accumulated over 8 years). They then switched to 0.62 threshold without acknowledging that there was anything wrong in using 0.6 (in fact just recently Met Lead on Facial Recognition, Lindsey Chiswick, stated that she was prepared to use even lower thresholds than 0.6 where the police’s own commissioned reseach shows it’s deeply biased – out of 33 falsely identified people, 30 were Black or Asian). And now (since 29th July) they have switched to 0.64, again without any acknowledgement of previous errors.
Systemic Racism
In must be stressed that our opposition to the Met Police’s use of live facial recognition surveillance is not due to its inaccuracies or bias, which do exist, but because it’s an oppressive instrument of control which targets the most marginalised members of our community, amongst them people of colour. In the hands of an institutionally racist police force it will inevitably produce racist outcomes. We witness this at this deployment.
Stops Log
Log Interval 14:30pm – 18:00pm
15.11 Stopped black person near van (2 black officers) no handcuffs whilst they checked his details. ID correct but data on him wrong – not wanted for anything, released after about 10 mins.
15.30 Stopped white person near market ( 1 black & 1 white officer) no handcuffs whilst they checked his details. Released after 10 minutes.
15.55 Stopped white person near van ( 2 white officers) no handcuffs whilst they checked his details. Released after maybe 5 mins.
15.58 Stopped white person after van (1 white and 1 black officer) no handcuffs whilst they checked his details. Arrest made, handcuffed and taken away.
16.14 Stopped black man near van (2 white officers) immediately handcuffed. Then started asking him questions, followed by checking his details. He offered no resistance, humiliated in front of his lady friend who was with him. After 15-20 mins released as he had no outstanding warrants against him – stopped in error.
16.30 Chased an alert down Church Lane towards East Croydon, but couldn’t find.
16.50 Stopped black man behind van (2 white officers) handcuffed immediately, whilst they checked his details. Kept him over 30mins in handcuffs. Humiliated as his friend walking by saw him in handcuffs. Twice we tried to intervene asking officers to take handcuffs off as he was not offering any resistance, they refused. Only when all checks came back negative that they finally released him.
Summary:
- 6 people stopped, only 1 arrested (83% stopped for no reason)
- Black innocent suspects on average stopped for between 18.3 to over 20 minutes.
- White innocent suspects on average stopped for 7.5 minutes.
- 2/3 Black suspects handcuffed when stopped whilst checks being done (even though all were innocent, not wanted for anything).
- No white suspects handcuffed when stopped whilst checks being done.
- When officers were white, black suspects were always handcuffed before checks were carried out
- When officers were black, black suspects like white suspects were not handcuffed.
The Stops Log shows that most black people who were stopped were immediately handcuffed, before any checks were carried out to confirm their identify and if they were wanted for anything. None of them had put up any resistance and were fully cooperating with the officers. There was no justification to handcuff them. All of them turned out to be innocent, not wanted for any crime, stopped because of police mistakes, handcuffed, it would seem, because they were black . None of the white suspects, on the other hand, were handcuffed whilst checks were carried out. How can the Met police explain this disparity between how black suspects are treated compared to white suspects?
It’s also interesting to see that it was invariably white officers who handcuffed black suspects for no reason.
Black suspects were also stopped for much longer, for at least twice as long as their white counterparts, sometimes over three times as long. Why? Do police database searches take longer if searching for a black person? Or is it the case that once they have shackled a black man there are in no rush to release him?
One black man, stopped and immediately handcuffed, was kept handcuffed for 15 to 20 minutes whilst they checked his ID and if he was wanted for anything. He was with a lady friend, he was not a flight risk, he didn’t resist, he fully cooperated with their demands – there was no reason to handcuff him. His friend was naturally embarrassed, she kept her distance. Even after the police admitted it was their mistake, and released him, when we talked to him she walked ahead, not wanting to be implicated in his stop.
This was not an isolated incident, it just gets worse…
Black Man Shackled For Over 30mins
One black man was handcuffed for over 30 minutes for no reason. We tried twice, unsuccessfully, to intervene on his behalf, explaining to the officers that they had no justification for keeping him in handcuffs. He was not resisting, it seemed it was purely to humiliate him. At one stage a friend of his happened to be passing by and saw him in cuffs. The man tried to explain “the police have got it wrong..” but his friend, perhaps worried they might also get arrested, kept distance and walked on.
The question arises, what was the purpose of handcuffing him for so long? Clearly he wasn’t a flight risk and had fully cooperated with the officers from the start, and from their body language it was clear the officers didn’t regard him as a threat. We wondered if there was an element, even subconsciously, of parading him to the public like a trophy that they had bagged a bad person, a dangerous black man, not dissimilar to the way runaway slaves were paraded after being caught to reassure the public of their safety. Certainly he caught the attention of many people walking past, a few pointing to him. For over half an hour he was paraded on Croydon’s busiest street, over a thousand people saw a black criminal in handcuffs, but very few saw an innocent black man being released by the police.
Is LFR Effective At Fighting Crime?
The police sells its intrusive surveillance of our community by fear, that without LFR technology, monsters – violent knife attackers, sex offenders and women beaters – will roam free on our streets. We did come across a couple of people who had bought into this narrative and said they felt safer. One even came to the police’s defence when an officer was being berated by a concerned citizen who wasn’t happy with his face being scanned. The reality is that the police’s own research – 29 different studies – show CCTV has no impact on violent crime – zero deterrence, and 33 studies which show that it does not reduce crime in city centres (College of Policing, Oct 2021).
The handcuffed black man, shackled for over 30mins, couldn’t understand why the police had stopped him in the first place, he told them “you have my address, if you think I did something wrong why didn’t you just come and knock on my door?” He had a point, if he was a wanted man with a warrant out for his arrest, why didn’t the police simply visit his home and arrest him? Wouldn’t that be more effective than waiting for him to walk through one of their biometric checkpoints? This raises the question of what purpose does LFR serve, does it actually keep criminals off our streets?
When I was going home at 6pm, as they were unmounting the last cameras from the van, I was walking along North End towards West Croydon station, about 100 yards from the deployment, when I notice two of the officers who were at the deployment just five minutes ago, now back on the beat, had arrested someone. That was one arrest by two officers in just 5 minutes on the beat, compare that to 15 officers taking 3.5h and the biometric checkpoint resulting in just one arrest. The case for LFR isn’t compelling.
About Control
But facial recognition surveillance was never about catching criminals, its about control through intrusive surveillance.
One person who stopped to talk to us, commented that facial recognition cameras could be made more discrete, that the real reason for the vans and giant cameras was deterrence – a show of force. There is something to that, the LFR van with its array of menacing cameras do have an imposing presence. It can be distressing even for a child. We saw one very young girl, needing no prompting from her mother, pulled up her polo neck over her face when she saw the cameras pointing down at her. An hour later, after finishing her shopping, when she returned with her mother, she was more defiant. This time shooting the cameras with her fingers. Perhaps there is a lesson in that, you cannot police through fear, it leads to resentment, defiance and resistance.
Policing Without Consent
The Cassey report concluded that the Met is no longer policing through the consent of the community, this is particularly true when it comes to live facial recognition. Three London boroughs have resoundingly rejected the police use of live facial recognition in their boroughs – Islington, Newham, and Haringey, all passing motions in their local councils banning the use of LFR by the police. Yet the Met police have ignored the wishes of the communities and still rolled out their LFR surveillance vans into these communities without their consent. This does not bode well for the future of policing.
Update - Deployment Log
Since writing this account, the Met Police have updated their deployment log to include this deployment. It shows that the deployment duration was 4h 50m, so we missed 1h 20m of it. Apparently 8 alerts were raised in total from which 3 arrests were made, they say ‘no action’ was taken for the remaining 5 – the reality is that action was taken – the 5 were stopped for no reason – some hand cuffed for over half an hour.
We witnessed 6 alerts, resulting in 1 arrest, so presumably in the 1h 20m we missed they must have arrested 2 people. Even with these extra arrests, their figures reveal that 62.5% of the time they got it wrong, stopping someone for no reason.
The log shows the Met estimate that 10,455 people walked past their live facial recognition van in the 4h 50m, that’s over a 1,000 people every 30 minutes which is how many people we estimated saw the innocent black man in handcuffs, humiliated, paraded like a criminal for over 30mins.
Update - Met Statement
A statement by Croydon Met Superintendent Mitch Carr on the LFR deployment, published on Croydon Communities Consortium website, reveals that during the deployment they arrested “A man who was seen to act suspiciously when he spotted the LFR van – he was found in possession of Class A drugs (cocaine.)” This confirms they had plain cloths officers stopping people who didn’t want to walk through the LFR system even though their sign states “there is no legal requirement for you to walk through the LFR system”.
Also interestingly Superintendent Mitch Carr reveals that Dr Jyoti Belur from the London Police Ethics Panel attended the deployment:
“We were pleased to be joined by Dr Jyoti Belur from the London Police Ethics Panel for Wednesday’s deployment. The LPEP is an independent panel set up by the Mayor of London to provide ethical advice on policing issues that may impact on public confidence. This is part of our commitment to operate the LFR technology as transparently as possible and in line with our principles of “policing by consent.” The officers are regularly receiving positive feedback from members of the public and while I respect the fact that there will be mixed views on this, I do believe we have good levels of approval from both partners and the public to utilise this technology to detect and deter crime in Croydon.”
Unfortunately she must have left before we arrived so didn’t get an opportunity to talk to her. We wonder if her presence was due in any part to the letter the community sent opposing the use of live facial recognition surveillance of our community?
It was interesting that one of the officers at the deployment on learning we were from the Resistance Kitchen reacted, without any prompting, “we have received the letter”. Whilst we have not yet received any reply to the letter which was addressed to the top echelons of the Met police, its clear it had an impact as officers at the deployment were briefed on it. We look forward to receiving a meaningful reply, hopefully the Met police will take this opportunity to listen to the concerns of the community.