Oxford Street Live Facial Recognition Cameras [2026]

Update — November 4, 2026: The Home Office has announced a new law and an independent oversight body able to pause police deployments. See UK police facial recognition law and oversight body.

London's Metropolitan Police is preparing to turn Oxford Street live facial recognition from an occasional van deployment into permanent infrastructure. According to a report published on October 24, 2026 by Reclaim The Net, citing Press Association reporting, fixed cameras are likely to be installed in November, in time for the Christmas shopping period. About 500,000 people walk the street every day, and every face that passes a live camera is compared against a police watchlist.

The plan first surfaced in summer 2026, when Mayor Sadiq Khan discussed it at a London Assembly meeting in July, and ITV News reported on August 12 that it would be among the first permanent installations in the country. What is new is the reported timeline, and the fact that the legal framework the government promised has still not arrived.

This guide explains what is planned, what the Met says about deletion and safeguards, what the numbers do and do not show, what the law actually says, and what you can do today, including how to protect bystanders' faces in any footage you publish.

Update — October 30, 2026: Suffolk police will run live facial recognition on November 3-5 and Hampshire reports 406,001 faces scanned in a year; see the UK live facial recognition rollout.

Update — October 27, 2026: Private retail is adding facial recognition too: see Facewatch Live, a free retail crime platform that can run alongside live facial recognition.

What Exactly Is Planned for Oxford Street?

The Met plans fixed live facial recognition (LFR) cameras on Oxford Street, probably attached to lampposts and similar street furniture rather than a police van. ITV News reported it would be among the first areas in the country with permanent LFR infrastructure. The Met has also discussed static cameras elsewhere in the West End and Soho, and the precise positions on Oxford Street were still being worked out as of October 24.

Mayor Khan said the Met is working with the Oxford Street Development Corporation on "how we use LFR in a sensible way." The street is being pedestrianised, with traffic removed over time, which is part of why the cameras can sit at pavement level without a vehicle in the way.

Police say the cameras will be switched on only when officers are deployed nearby, and that signs will tell people when a deployment is active. In other words, the hardware is permanent but the scanning is described as intermittent. That distinction matters, because a camera network that is installed and wired up is far easier to switch on more often than a van that has to be driven to a site.

Commissioner Sir Mark Rowley has called live facial recognition "one of the most revolutionary technology advances in policing in recent years." Supporters in politics include Conservative leader Kemi Badenoch, who has said the party "will expand live facial recognition," and Policing Minister Sarah Jones, who said "Law abiding citizens have nothing to fear."

How Does Live Facial Recognition Actually Work?

Live facial recognition detects every face in a camera's view, converts each into a numerical template, and compares it against a watchlist of people police want to find. If the similarity score passes a threshold, the system raises an alert and an officer decides whether to stop the person.

The important point is that the system must process everyone, not just suspects, in order to find the few it is looking for. A watchlist match is the output; the input is the face of every shopper, tourist, commuter and child who walks through the camera's field of view. That is why privacy groups describe it as biometric identification at scale rather than as ordinary CCTV.

Compared with a traditional camera, three things differ:

• It creates biometric data. A faceprint is a unique identifier derived from your body, closer to a fingerprint than to a photo.
• It is automatic. Nobody has to watch the screen for the identification to happen.
• It is silent. Unlike a fingerprint scan, you do not have to cooperate or even know it occurred.

For the same technology in a different setting, see how the Home Office has described its own scheme in our guide to the UK Home Office live facial recognition DPIA, and how Western Australia's trial results compared in WA Police's live facial recognition trial.

What Does the Met Say About Deletion and Safeguards?

The Met says data from people who do not generate an alert is deleted immediately, that watchlists are deleted within 24 hours, and that footage from LFR CCTV is deleted within 31 days. These are the force's own assurances as relayed by Reclaim The Net, and they are policy statements rather than independently audited facts.

It is worth being precise about what each retention period covers:

The last row is the one people tend to miss. Even if templates vanish instantly, the underlying video of your face can remain for up to 31 days. Whether that footage can be re-run through a face matcher later is a policy question, not a technical impossibility, and is one of the areas where a proper statutory framework would normally set limits.

Is Live Facial Recognition Legal in the UK?

There is no law written specifically for it. Reclaim The Net reports there is no legal framework specific to LFR, and Liberty's Akiko Hart says shoppers will be scanned by "unregulated surveillance technology without being given a choice." Police instead rely on a patchwork of common law policing powers, data protection law and their own published policies.

Two pieces of context are essential:

1. The consultation is unanswered. The Home Office launched a consultation in December 2025 on a legal framework for police biometrics, including LFR. It closed in February 2026, and as of October 24 no response had been published.
2. The courts have not blocked it. In April 2026 the High Court dismissed a challenge brought by a youth worker who was wrongly flagged outside London Bridge station, together with Big Brother Watch.

For the broader data protection backdrop, our explainer on UK GDPR and the Data Protection Act for video privacy sets out why biometric data used to identify a person is classed as special category data, and our coverage of the Clearview AI UK tribunal and Court of Appeal fight shows how the regulator has approached private-sector face matching.

Do the Numbers Show It Works?

The numbers are disputed, and which ones you quote depends on what you want to prove. The Met says it has attributed more than 2,000 arrests to LFR since the start of 2024, and that fixed cameras in Croydon "contributed to a 10 per cent reduction in crime" and led to 173 arrests.

Critics cite other figures from the same trials. In Croydon, of 470,000 people processed, 99.96 percent had nothing to do with any crime, according to the figures quoted by Reclaim The Net. A British Transport Police trial in a station scanned 926,912 faces and produced zero arrests.

How to read this fairly:

• For: If a small number of dangerous people are caught who would otherwise have escaped, supporters argue the scan of everyone else is a modest cost, especially if non-matching data is deleted at once.
• Against: Opponents argue that scanning hundreds of thousands to find a few is the definition of indiscriminate surveillance, and that a low false alert rate still means real people get stopped. The youth worker in the High Court case was one of them.
• Missing: Crime reduction claims are hard to attribute to one tool when other policing changes are happening at the same time, and the Met's figures were presented without an independent evaluation in the coverage we reviewed.

Reclaim The Net is an advocacy outlet and its framing is critical; the police figures above are the Met's claims, and the critics' figures are as reported by that outlet. Treat both as claims to check against the original publications.

What Are Privacy Groups Saying?

Big Brother Watch and Liberty oppose the plan, and say the debate should happen in Parliament rather than be settled by installation. In an August 12 statement, Big Brother Watch's Senior Legal and Policy Officer Jasleen Chaggar said:

"Innocent Londoners who are not suspected of any crime by the police should be able to wander around public spaces without identifying themselves to the authorities."

She added:

"A face scan is as unique as a fingerprint. The police cannot take your fingerprints when you are not suspected of wrongdoing. Taking a faceprint is no less intrusive just because it happens remotely."

The fingerprint comparison is the heart of the civil liberties argument. It frames a faceprint as a bodily identifier that should require suspicion before it is taken. The counter-argument from police and ministers is that a face is already visible in public, and that a scan which is deleted within moments for non-matches is not comparable to retaining a fingerprint.

ITV's August report also cited Liberty-commissioned polling, which found 48 percent of people oppose scanning the faces of those walking on high streets when there is no suspected imminent threat. That is a poll commissioned by a campaigning group, so read it as one data point rather than a settled measure of opinion.

Why Does a Permanent Installation Matter More Than a Van?

A permanent network changes the default from "police occasionally scan a street" to "the street is instrumented and scanning is a software setting." Three practical differences follow.

Cost and friction fall. A van needs officers, a location, a plan and a sign each time. Fixed cameras already wired to the system reduce the cost of each additional deployment, which makes frequent use easier.

Coverage becomes habitual. Oxford Street is a route for commuters and workers as well as shoppers. People who pass daily will be scanned repeatedly whenever it is switched on, which is different from a one-off encounter.

Function creep is a policy risk, not a certainty. Cameras installed for a watchlist of wanted offenders can in principle be pointed at other lists. Whether that happens depends on the legal limits that do not yet exist, which is why the unpublished Home Office response matters so much.

Similar concerns drove public debate when France's police used facial recognition on social media images after riots, and when Meta's smart glasses prompted ACLU backlash: once the capability is normal, the argument shifts from whether to how often.

Can I Avoid Being Scanned on Oxford Street?

Only by not being in the camera's view while it is active. There is no opt-out, no consent step and no way to decline at the point of scanning. Masks, hats and sunglasses can reduce match quality, but they are not reliable and officers can ask for a face covering to be removed in some circumstances, so we would not present any of them as a guaranteed defence. Our investigation into whether digital camouflage defeats face detection found that tricks which fool one model often fail on the next.

What you can do is take the practical steps below, and exercise your rights if you are stopped.

What Should You Do Today?

1. Read the signs. The Met says signs will show when a deployment is active. If you do not want to pass through, take another route.
2. Know what happens after an alert. If an officer stops you because of a match, you can ask why you were stopped and what the basis for the match was. If you are wrongly flagged, note the time, place and officer details and contact Big Brother Watch or Liberty, both of whom have supported people in that position.
3. Respond to the Home Office. When the response to the consultation is published, check whether it creates statutory limits on watchlists, retention and audit. Contact your MP if it does not.
4. Ask for your data. Under UK data protection law you can make a subject access request to the Met for personal data it holds about you, including footage in which you appear.
5. Blur bystanders before you publish. If you film on a busy street, assume every face in the clip is identifiable. Blur faces before you upload, particularly children's.
6. Check old footage. Search your own channels and cloud albums for clips of crowds that you would not publish today, and re-export them with faces blurred.
7. Do not rely on weak blur. Light pixelation can sometimes be reversed. Our guide to forensic de-pixelation and why weak blur fails explains what strength is needed.

How Can You Protect Faces in Video You Publish?

You cannot control police cameras, but you control what you share, and that is where face search engines and scrapers get their material. Footage from a busy shopping street shows dozens of faces, and each one is a potential match for tools like those discussed in our explainer on whether AI can find exact matches of faces across websites.

BGBlur is a browser-based tool that detects faces in video and photos and blurs them with motion tracking, so you do not have to edit frame by frame. The basic workflow takes three steps:

1. Upload your clip or photo in the browser.
2. Select faces to blur. Detection finds them automatically, and you choose which to keep clear, for example your own.
3. Export the blurred file in MP4, MOV or WebM.

Our guide to blurring faces in photos and videos with AI walks through it, and the street photography privacy guide compares approaches for people who shoot in public.

An honest caveat: blurring your own footage does nothing about a police camera, a shop CCTV system or footage someone else uploads. It protects the people in what you publish. Check results before posting, because small or distant faces and heavy motion can need a manual pass, and BGBlur's free plan has limits (200MB, 10 minutes, 3 videos a month, 720p export) that may matter for long street clips.

Who Is Affected Beyond Shoppers?

Workers, residents, visitors and anyone filming in the area are all in range. Retail staff who stand at shop doors on Oxford Street will be scanned many times a day, which raises employment questions the current guidance does not address. Street performers, journalists and activists have particular reasons to care about being identified in public. Parents with children in prams are scanned too, and the Met's deletion pledge is the only protection stated for their data.

Creators who film vlogs, interviews or news segments on the street also face a publishing decision. Showing recognisable faces of passers-by is lawful in many contexts in the UK, but it still creates identifiable personal data about people who did not agree to it. Our guide on unauthorised filming and face blurring covers when to blur even where you are not strictly required to.

What Happens Next?

Three things to watch over the coming weeks: whether the cameras go up in November as reported; whether the Home Office finally publishes its consultation response and proposes statutory limits; and whether further legal challenges are brought once a permanent installation exists, since a fixed network gives claimants a concrete target that van deployments did not.

ITV reported in August that the technology had also started to be deployed at London's Tube stations, so the outcomes of those trials may shape the debate. We will update this post if the Met announces switch-on dates or the Home Office publishes its response.

Conclusion: Plan for a World Where Public Faces Are Scanned

Oxford Street's cameras are not a done deal until they are switched on, but the direction is clear: live facial recognition is moving from a trial to a fixture of one of Britain's busiest streets, without the dedicated law that campaigners say should come first. Supporters point to arrests and a drop in crime; critics point to hundreds of thousands of innocent people scanned and trials that produced nothing. Both sides agree the technology is arriving faster than the rules.

You cannot opt out of a police camera, but you are not powerless. Know your rights if you are stopped, respond when the Home Office publishes its framework, and be deliberate about the faces in the footage you publish. If you film in public, blur bystanders before you post, with a method strong enough to hold up. Try BGBlur's face blur in your browser on your next street clip, and keep an eye on this story as the November installation date approaches.