TI-2026-061O — Just Public

Classification: HIGH · Biometric Surveillance · Face Recognition / "Open Data" Confidence: HIGH on the Clearview record and the regulatory rulings (public reporting, official orders); MEDIUM on the consumer-face-search generalisation. TLP:WHITE — public reporting, public regulatory decisions, public court settlements. Series: The Fake Umbrella — Part O. O is for "Open data." Two words that turn the harvesting of your face into a shrug. It was already public, so what could be wrong with collecting all of it, forever, and making it searchable? Facts are facts. They are good for some and bad for others. You decide.

The seed: a photo of a stranger becomes their name

Imagine you take a candid photo of a person you have never met — on a train, at a protest, across a café — and within seconds you have their name, their employer, their old dating profile, the town they grew up in, and every other place their face has ever appeared online. That is not a hypothetical. It is the product, and the company that made it famous is Clearview AI.

Clearview, exposed by Kashmir Hill in The New York Times in 2020, built a facial-recognition engine by scraping billions of photos — from Facebook, Instagram, LinkedIn, Venmo, and the open web — without anyone's consent, and turned them into a searchable biometric index of humanity. Upload a face; get the person. It has claimed a database in the tens of billions of images, and it sold access to police forces and agencies around the world. For the first time, a face in a crowd is a search query, and the person behind it has no idea they have been looked up.

And when asked how this could possibly be allowed, the company gave the answer that is the whole of Part O: "It's just public information. These photos were already online."

Why "already public" is the wrong frame — three ways

The "just public" defence sounds unanswerable until you take it apart. It fails on three separate, independent grounds — the same aggregation problem this series has met before (Parts D and G), now applied to the one datum you can never change.

1. Aggregation is a new capability, not a mirror. A single public photo on a friend's feed and a searchable, worldwide, biometric index that resolves any face to an identity are not the same object at different scales — they are different objects. The first is a picture; the second is a manufactured power that did not exist before someone built it. "It was already public" describes the raw material; it says nothing about the machine, and the machine is the harm. As with the location brokers of Part D, the aggregation is the reason there is a product and a price.

2. A face is not a password. This is the point that makes biometrics uniquely grave. When a password leaks, you change it. When a credit card is stolen, you cancel it. Your face cannot be reset. It is a permanent, unique, lifelong identifier that you wear in public every day and cannot revoke, rotate, or hide without ceasing to live a normal life. To index it into a searchable surveillance key is to inflict a harm that is irreversible — you can never make yourself un-searchable again. "Just public" quietly treats a face like a tweet. It is closer to a fingerprint you can never take off.

3. It was ruled illegal — repeatedly, by democracies. This is not a fringe civil-liberties complaint; it is the settled finding of regulators. Clearview has been fined and ordered to delete data by the UK's ICO, France's CNIL, Italy's Garante, and the authorities of Australia and Canada, all for unlawful biometric processing without consent; and in the US it settled the ACLU's BIPA case (Illinois Biometric Information Privacy Act, 2022), restricting its sales to most private companies. Multiple functioning democracies looked at "it's just public data" and formally answered: no, it isn't lawful.

The thing it actually kills

Step back from Clearview the company to the capability itself, because Clearview is only the first to get caught doing it at scale — the consumer version, PimEyes and its kind, already lets anyone pay to find where a face appears online. What this capability abolishes is not a data point. It is anonymity in public — a freedom so old and so ambient that we never named it.

For all of human history, being a face in a crowd meant something: you could attend a protest, walk into a clinic, meet a lover, leave an abuser, or simply exist in public without every stranger being able to instantly know who you are and everything about you. That practical anonymity is the substrate of a free public life. Face search deletes it. A cop can identify every protester from a photo. A stalker can turn a glimpse into a home address. A hostile government can put a name to every face in a demonstration. An abusive ex can find where you moved. The "open data" was the material; the product is the end of being no one in particular, and it was built and sold before almost anyone got to decide whether they wanted to keep that freedom.

The seven questions

Who is this? Clearview AI and the face-recognition industry (including consumer face-search like PimEyes) — firms that turned scraped public photos into a searchable biometric identity engine.

Why did it attack? No attack — a harvest. It scraped the world's faces to build and sell an identity-resolution product to police, agencies, and (in the consumer tier) anyone.

Who does it claim to be? A tool using "publicly available information" to help solve crimes — neutral use of data that was "already out there."

Whom is it related to? The observatories (061G — aggregating scattered public data into a searchable index), the academic face-recognition pipeline (061K — whose technique it productised), and the "commercially/publicly available" defence of the location brokers (061D).

Where is the money? Contracts to police and agencies for search access, and consumer subscriptions to face-search. The indexed person is the merchandise, never the customer, and was never asked.

What is the official anchor? None that survived scrutiny — the anchor was "it's public," and regulators across five-plus democracies rejected it as unlawful biometric processing.

What if it is not who it claims? It is who it claims — a face-search company using public photos. The finding is that "using public photos" at this scale is the construction of a non-consensual, irreversible, anonymity-ending biometric surveillance system.

What if…

• What if it helps catch criminals? It can — and that is the fig leaf, not the refutation. A capability that resolves any face to an identity is defined by its universality, not its best case; the same search that identifies a suspect identifies a protester, a source, a survivor. Utility to police does not answer for building a searchable face-print of everyone who ever posted a photo.
• What if I have nothing to hide? Anonymity in public is not about hiding wrongdoing; it is the ordinary freedom to move through the world un-catalogued — to grieve, protest, seek help, or simply exist without a stranger instantly knowing your name. "Nothing to hide" mistakes a structural freedom for a personal secret.
• What if the photos really were public? They were — and a face is not a password, so "public" here means "permanently and irreversibly capturable." The publicness of the raw photo is exactly what makes the harm total: you cannot un-publish your own face.

Steelman: the strongest defence — and why it fails

Built at full strength:

"Clearview uses only publicly available images — the same ones any person or search engine can see. It has helped police identify child abusers, fraudsters and violent criminals. Publishing a photo publicly is a choice; indexing public information is what search engines do; and free-speech/‑information principles protect collecting what is openly posted. Banning it would cripple legitimate investigations over a theoretical privacy concern."

Where it breaks:

• "Search engines index public info" conflates text with biometrics. A web index points to documents; a face index converts a person's body into the query. Google finding a page you wrote is not the same as a machine turning your face, involuntarily, into a lifelong lookup key — and the law of five democracies has drawn exactly that line.
• "Helped catch criminals" is the universal-tool fallacy. Every mass-surveillance capability can point to a caught criminal; that never justified the capability's reach over everyone else. The question is not "did it ever help," but "should a searchable biometric index of all humanity exist," and the harm is borne by the billions who are not suspects.
• "Public posting is consent" is precisely what regulators rejected. Posting a photo to friends is not consent to have your biometric face-print extracted, stored, and sold to police forever — the UK, French, Italian, Australian and Canadian authorities and the BIPA settlement all found that no such consent existed.
• The one true thing: narrow, consented, warranted, tightly-overseen facial recognition — a person unlocking their own phone, a properly-authorised one-to-one match under law — can be legitimate, and this dossier does not deny it. But that is a specific, bounded act. Scraping the world's faces without consent into a searchable index sold to whoever pays is not that act; it is the abolition of public anonymity dressed as "using public data."

Verdict: narrow, consented biometric use is a real debate. Face-scraping "open data" surveillance as built — non-consensual, irreversible, anonymity-ending, and ruled unlawful across multiple democracies — is HIGH. "Just public" is not a mitigation. It is the two-word shrug that let someone build a searchable face-print of humanity before anyone was asked. Confidence: HIGH on Clearview and the rulings; MEDIUM on the consumer generalisation.

Read between the lines

The move in "open data" is to relocate the decision into the past and out of your hands. By saying the photo was already public, the company quietly asserts that the moment you or a friend posted a picture, you consented to every future use of your face anyone could ever invent — including uses that did not exist and could not be imagined at the time. It is consent retroactively manufactured out of a birthday photo from 2011. And notice what "public" is made to carry: in ordinary life "public" meant visible to the people around you in a moment — a fundamentally forgetful, local, human kind of exposure. Face search redefines "public" as permanently, globally, machine-searchable — a total and eternal kind of exposure that no one ever agreed to, because it never existed to agree to. The two words hide an enormous swap: they take the old, safe, forgetful "public" you actually lived in, and hand you back a new, total, permanent one, and call it the same thing.

The human layer

A woman leaves a man who hurt her. She does everything right: new city, new job, no forwarding address, careful. And one day he takes a photo of a stranger who reminded him of her, or finds an old picture, or simply pays a few dollars to a face-search site — and the machine that scraped "just public data" returns the café she now works at, three towns away, because a colleague tagged her in a group photo she never saw. She did not post it. She did not consent. She could not have un-published her own face even if she had known. The freedom that protected her — the ordinary, ancient ability to be a new face in a new place — was quietly deleted, for everyone, so that a company could sell face-search to police and, in the consumer tier, to him. That is the human weight of "open data": not an abstraction about privacy, but a survivor found, a protester named, a source unmasked, because the one identifier a person can never change was harvested from the world and made searchable, and the harvest was excused with two soft words. Your face was always public in the old sense — visible to the room. It was never meant to be public in this one — visible, forever, to everyone with a photo and a login.

Defense — refuse the redefinition of "public"

(1) Name the swap. "Open data / publicly available" is doing enormous work — insist on the distinction between visible in a moment and permanently machine-searchable; the second is a new thing that requires new consent, and regulators agree. (2) Assume your face is indexed and act accordingly — high-risk people (survivors, activists, sources) should treat any public photo as potentially searchable, and platforms/allies should minimise tagging and public imagery of them. (3) Support and use the legal levers that worked — BIPA-style biometric-consent laws and the ICO/CNIL/Garante model are the tools that actually fined and constrained this; they are the proof that "just public" is defeatable. (4) For policymakers: biometric identifiers are irreversible and therefore deserve the strongest, opt-in, deletion-backed regime — the honest question is not "can this help an investigation" but "should a searchable face-print of every human exist for sale at all," and everything else is the euphemism doing your deciding for you.

Investigation metadata

Next — TI-2026-061P: P is for "Privacy" — the "privacy" VPNs and "no-log" services that promised to hide you and quietly logged, harvested, or handed over exactly what they swore they wouldn't.

"It's just public information" is the softest sentence in the whole umbrella and one of the most consequential, because it relocates a decision you never made into a birthday photo from years ago and calls it consent. Your face was public the way a face in a crowd is public — briefly, locally, forgettably. They scraped it into something else: permanent, global, searchable, for sale, and impossible to take back, because a face is not a password. The survivor found in a new town, the protester named from a crowd — that is what two soft words were hiding. Facts are facts. They are good for some and bad for others. You decide.