SWITCHED ON
The daily technology series nobody asked for but everyone needed
Smile, You're Being Catalogued: The Facial Recognition Reckoning
Having a face in public is not consent to be identified, tracked, and logged. And yet here we are — being scanned, matched, and filed away by systems most of us can't see and none of us agreed to.
Being visible is not the same as being identified. Being seen is not the same as being catalogued. The comparison is a category error dressed up as a legal argument.
In Episode 4 we established that your data is being collected, packaged, and sold by companies you've never heard of, that the legal right to make them delete it exists in theory and is largely toothless in practice, and that the tech industry has designed its own compliance tools to be just effortful enough that most people give up. If you read that and felt a low-grade, hard-to-place unease settling somewhere in your chest — good. That's the correct response. Hold onto it. Because today we're going somewhere that makes the data broker problem feel almost quaint.
Today we're talking about your face. Not metaphorically. Literally. The specific arrangement of geometry that constitutes your face — the distance between your eyes, the width of your nose, the shape of your jawline — is, in 2026, a piece of biometric data that can be captured from a distance, without your knowledge, without your consent, processed against a database of millions of faces in seconds, and used to identify, track, and log your movements through public space with an accuracy that would have seemed like science fiction fifteen years ago.
And it is happening. Right now. In shopping centres, airports, sports stadiums, casinos, schools, and on street corners in cities across the world. Sometimes disclosed. Often not. Regulated in some places. Completely unregulated in many more. Operated by governments, corporations, landlords, employers, and anyone else who can afford the software — which is becoming cheaper every year.
01 — What Facial Recognition Actually Is, and How Good It's Got
Facial recognition, in the context we're discussing, refers to automated systems that can identify a specific individual by comparing an image of their face against a database of known faces. The identification piece works roughly like this: a camera captures an image, software maps the geometry of the face and converts it into a mathematical representation — a faceprint — then compares that faceprint against a database. If there's a match above a certain confidence threshold, the system returns an identification. On modern hardware, this takes between one and three seconds.
The accuracy of leading commercial systems, under controlled conditions, is now extraordinary. Error rates for the best systems identifying adults in good lighting are in the low fractions of a percentage point. The headline numbers are impressive enough that it's easy to forget what "low fractions of a percentage point" means when you're running millions of identifications — you're generating thousands of false positives, many of which will be acted upon by someone, somewhere, with consequences for the person incorrectly identified.
Accuracy drops significantly when the subject is a person of colour, a woman, or an older person. A technology that is significantly more likely to misidentify a Black person than a white person, deployed to identify criminal suspects, is not a neutral tool. It is a bias machine with a badge on it.
02 — Where It's Being Used: A Brief and Unsettling Tour
Law enforcement is the most extensively documented use case. Police forces in the UK, US, China, and many other countries use facial recognition to identify suspects in CCTV footage, to run live identification on crowds at public events, and in some cases to generate lists of persons of interest. The Metropolitan Police in London has deployed live facial recognition at events and in specific areas of the city. Several US cities have banned police use of the technology following documented cases of wrongful arrests based on false matches — all of which, in the American cases, involved Black men.
Retail is the use case most people don't know is happening. Major retailers in multiple countries have deployed facial recognition to identify known shoplifters, monitor staff, and — more controversially — track the movements and behaviour of customers for commercial purposes. In most jurisdictions, the legal basis for this is murky at best. In many, it is simply assumed until challenged.
Transport hubs — airports, train stations, border crossings — are among the most extensively deployed environments globally. Airlines and border agencies use facial recognition to verify identity against passport databases, often framed as a convenience feature that also happens to be compulsory if you want to board. What is presented as frictionless travel is also, from a civil liberties perspective, the normalisation of biometric surveillance as the price of moving through public space.
Schools. Yes, schools. Multiple countries have piloted or deployed facial recognition in educational settings, ostensibly to track attendance and monitor who enters the building — and in some cases to assess student engagement or emotional state in classrooms. The deployment of emotion-reading AI systems on children in educational environments is, to put it as plainly as possible, one of the most ethically indefensible applications of this technology currently in operation anywhere.
03 — The Consent Problem, Which Is Actually the Everything Problem
Here's the fundamental issue with facial recognition in public spaces that tends to get lost in the more technical debates: it is, by its nature, a technology that operates without meaningful consent. When you walk past a camera equipped with facial recognition software, you have not agreed to be identified. You cannot practically refuse — short of not leaving your home, which is not a realistic exercise of a right. You cannot see the system operating. You cannot know what database your faceprint is being compared against, what the result is being used for, who it's being shared with, or how long it's being retained.
Being visible is not the same as being identified. Being seen is not the same as being catalogued. The ability of any individual to visually recognise a face they've seen before is not remotely comparable, in scale or capability, to automated systems that can identify millions of faces against global databases in real time.
The argument made by proponents — that you have no reasonable expectation of privacy in public space, that you can be seen by any passerby, that CCTV has existed for decades — misses the qualitative difference entirely. The comparison is a category error dressed up as a legal argument, and it has been repeated so many times by people who should know better that it has acquired an undeserved credibility.
04 — The Regulatory Landscape: A Study in Fragmentation
The EU's AI Act — the most comprehensive AI regulatory framework currently in force anywhere in the world — includes significant restrictions on real-time facial recognition in public spaces. Live remote biometric identification by law enforcement in public spaces is prohibited except in narrowly defined circumstances: searching for missing persons, preventing imminent terrorist attacks, identifying perpetrators of serious crimes. The framework is real, enforceable, and meaningfully more protective than what exists elsewhere.
In the United States, the picture is characteristically fragmented. There is no federal facial recognition law. A handful of cities have banned government use of the technology. Illinois's BIPA is the most robust state-level protection and has been the basis for significant litigation. But the majority of US jurisdictions have no meaningful regulation, law enforcement agencies operate with substantial discretion, and commercial deployment in retail and private spaces is almost entirely unregulated.
China is in a category of its own. The scale of facial recognition deployment by the Chinese state — integrated into a surveillance infrastructure that monitors movement, behaviour, and social compliance across urban environments — represents the most extensive deployment of the technology anywhere in the world. It is cited as a cautionary tale in Western regulatory debates and, less publicly, as a capability benchmark by security agencies in democracies that might prefer you didn't think too hard about their own ambitions in this area.
05 — The Deeper Question Nobody Wants to Answer
Here's where this gets genuinely philosophical and genuinely important. We need to decide, as societies, what kind of public space we want to live in. The optimistic case — made by security agencies, retailers, and technology companies with products to sell — is that facial recognition makes environments safer, reduces crime, and streamlines movement. These benefits are real and measurable in at least some contexts.
The pessimistic case — made by civil libertarians and anyone who has spent time thinking about the historical relationship between surveillance capability and state power — is that a world in which your movements through public space can be tracked, recorded, and associated with your identity in real time is a world in which the practical exercise of freedoms like protest, political dissent, and anonymous assembly becomes significantly more difficult. Not impossible. Not immediately suppressed. Just chilled. The knowledge that you are being watched changes behaviour. It is supposed to. That is, often, the point.
The question of whether that trade — some increment of safety and convenience for a significant reduction in practical anonymity — is worth making is not a technical question. It is a political and philosophical one. And it is being made right now, without most people's knowledge or meaningful participation, by governments and corporations and landlords and school administrators who have concluded, largely without asking, that the answer is yes. That conclusion deserves to be challenged. Loudly, persistently, and before the infrastructure of ubiquitous facial recognition becomes sufficiently entrenched that the challenge becomes academic.
Tomorrow we're going somewhere that might feel like a relief after five consecutive days of surveillance and privacy erosion: the metaverse. Specifically whether the concept that was going to redefine human connection and justify hundreds of billions in corporate investment is, in 2026, a transformative emerging reality or the most expensive lesson in the history of tech hubris. The answer is more interesting than you might expect. See you then.
Switched On is a daily technology series covering AI, social media, data privacy, and the digital forces reshaping modern life — with no corporate spin, no false comfort, and absolutely no mercy for buzzwords.
