UK Facial Recognition Gamble: Safety first, science later?

Is it “bias” if a camera wrongly flags a suspect – or just the cost of catching real criminals? And how many murders, stabbings and rapes should Londoners accept every August Bank Holiday in the name of culture and tradition? As the Met prepares to unleash live facial recognition at Notting Hill Carnival, the clash between cultural pride, political language-play, and public safety is impossible to ignore.

Every August Bank Holiday, Notting Hill becomes Europe’s biggest street party – and one of London’s most demanding policing operations. This year the Metropolitan Police will deploy live facial recognition (LFR) on a scale never seen before. The promise is bold: a technology that can pick out criminals in real time without racial, gender or age bias. The reality is more complicated. The science is contested, the stakes are high, and the backdrop is a festival that now records murders, stabbings and assaults with grim regularity.

What is LFR?

Live facial recognition involves scanning crowds in real time and comparing faces against police watchlists. Unlike CCTV, which simply records, LFR is proactive: it flags suspects on the spot. The technology is controversial because it can misidentify people, and those errors are not random. Studies worldwide have shown higher error rates for women, older people and ethnic minorities cos they all look similar.

The study at the centre of the row

The Met leans heavily on a 2023 study by the National Physical Laboratory. It claims that at a sensitivity setting of 0.64 the system produces no statistically significant bias. Yet Professor Pete Fussey, one of the UK’s leading experts on surveillance, has pointed out that the conclusions rest on a vanishingly small sample: just seven false matches. The study involved 400 volunteers over 34.5 hours. In policing terms, that is a drop in the ocean.

The Unfair “Bias” Narrative: The trade-off no one admits

Facial recognition settings matter. Lower thresholds cast the net wide, flagging more potential suspects but inevitably creating more false positives. Critics call this “bias” because those errors often fall disproportionately on women, older people and minorities.

But here is the difference. When we talk about heart disease, cancer screening or white-collar fraud, false positives are accepted as the cost of detection. No one suggests a blood test is “biased” against men because it produces more false positives for them. No one worries that an accounting audit might be “unfair” to bankers because it flags some clean transactions along with the dodgy ones.

But ironically, when the conversation shifts to street crime – to potential murders, robberies, and religiously-motivated terrorist plots – the language suddenly changes. We no longer talk about margins of error, we talk about “bias.” We move from statistics into moral philosophy. Is this science, or is it word-play designed to paralyse action?

Higher thresholds reduce false positives but increase the chance of missing genuine suspects. It means genuine robberies, genuine murders and genuine rapes going undetected. The Metropolitan Police insist they have found the sweet spot at which the system avoids bias. The truth is that this “bias-free” claim has never been tested at the scale London (and the UK as a whole) demands.

The bigger picture

Londoners face real threats. As reported by The Guardian, before the carnival, the Met seized 11 firearms and more than 40 knives, and arrested over 100 suspects. LFR promises efficiency, deterrence and speed in such environments. But it also hands unprecedented power to algorithms that remain poorly understood and lightly scrutinised.

So the dilemma is not whether LFR works. It clearly does. The question is how much uncertainty society is prepared to accept in the name of safety. Are Londoners comfortable trading untested claims of neutrality for a promise of security? And why, once again, is the burden of proof placed on critics rather than on the state itself? There are valid questions and arguments on both sides.

The Crime Question the Notting Hill Carnival Cannot Ignore

Notting Hill Carnival is Europe’s largest street festival, drawing up to two million visitors into narrow residential streets over a single Bank Holiday weekend. The scale is extraordinary. So are the policing demands and the level of serious crime that now shadows the event.

The record is clear.

• In 2024 there were two murders and seven non-fatal stabbings.
• Police made 349 arrests, including 72 for offensive weapons, 13 for sexual offences, 53 assaults on emergency workers and 61 assaults on police officers.
• In the run-up to Carnival 2025 the Met conducted intelligence-led operations that resulted in 100 arrests, 21 people recalled to prison, and 266 people barred from attending. Officers seized 11 firearms and more than 40 knives.
• By Sunday evening of the weekend itself, police reported 140 arrests, including 13 linked to live facial recognition alerts on the approaches.

What residents report.

Local households and traders describe the weekend as a lock-in. Many say they board up properties to protect against looting and vandalism, some businesses close, and community groups speak of streets feeling overwhelmed. These are perceptions, not formal statistics, but they are widespread and consistent.

What the police response tells us.

Nearly 7,000 officers were deployed this year, which is the largest contingent yet. That number is not theatre. It is an admission that, without extraordinary resources, the event risks becoming unmanageable. Symptoms.

What the media tend to gloss over: Context without excuses.

On paper, Carnival’s arrest rate looks similar to other large gatherings. But the comparison is misleading. Glastonbury, Reading or Wembley rarely see murders or multiple stabbings in a single weekend. Arrests for firearms possession at those events are virtually unheard of. Sexual assaults are far less frequent.

Notting Hill Carnival is different. The concentration of serious violence – murders, knife attacks, rapes, assaults on police, and looting – is unique, and it happens within the confines of a dense residential neighbourhood. That is the problem policymakers must solve. It is also the uncomfortable truth much of the media and intellectual elite prefer not to discuss.

The policy question.

So, if the event remains in its current location, what level of risk is the community expected to absorb each year. If it relocates or introduces tighter controls, what cost does that impose on authenticity and access. Culture matters. But, so does safety. The choice is not symbolic or political; it should be factual. It is measured in arrests, hospital admissions and the number of officers required to hold the line. Facts.

The hard choice ahead

In the end, the question is not whether live facial recognition works, but whether London and the UK as a whole is willing to face the truth about the Carnival itself. A celebration of culture, yes – but also a magnet for knives, guns and assaults. Policymakers argue and can tinker with algorithms all they like, yet the real decision is starker: how much violence are we prepared to tolerate in the name of culture and tradition without taking a deeper look?

Stay ahead of the headlines. Follow our UK News section for the latest crime, policing and politics stories shaping Britain today.

And, for headlines that roam across the Eurozone, try our European news section.