A look at Clearview AI, a facial recognition app claiming it scraped 3B+ images from sites like Facebook, YouTube, and that 600+ law enforcement agencies use it
A little-known start-up helps law enforcement match photos of unknown people to their online images — and “might lead to a dystopian future or something,” a backer says. Tweets: @kashhill , @nytimes , @bernstein , @davidenrich , @jorge_guajardo , @saralovesyou , @davidplouffe , @nytimes , @mattblaze , @jameeljaffer , @jasonschreier , @kashhill , @kashhill , @kashhill , and @kashhill Tweets: Kashmir Hill / @kashhill : The privacy paranoid among us have long worried that all of our online photos would be scraped to create a universal face recognition app. My friends, it happened and it's here: https://www.nytimes.com/... @nytimes : EXCLUSIVE: With a single photo, hundreds of law enforcement agencies in the U.S. can learn your name, where you live, and who you know. It will end your ability to walk down the street anonymously. https://www.nytimes.com/... Joe Bernstein / @bernstein : A major takeaway from @kashhill's incredible story. Ethically-dubious companies have had the ability to create facial recognition tech on a terrifyingly vast scale for years. They haven't done so because even they recognized it was too evil https://www.nytimes.com/... pic.twitter.com/klXLlzb9Ue David Enrich / @davidenrich : Exclusive: Hundreds of law enforcement agencies (incl @FBI and @DHSgov) have quietly started using a facial-recognition program that ends your ability to walk down the street anonymously. It's funded by @peterthiel and has links to alt-right. by @kashhill https://www.nytimes.com/... Jorge Guajardo / @jorge_guajardo : File under things that would be scandalous if they happened in China, instead of the USA. https://www.nytimes.com/... @saralovesyou : We really do need to have a government that deals with this. https://twitter.com/... https://twitter.com/... David Plouffe / @davidplouffe : Let's hope in future debates, candidates are asked about this, AI, changes in warfare, different education models. Much more important than the same old exchange on health care. These are all things that will or are happening and will greatly affect us all. https://twitter.com/... @nytimes : .@kashhill got to see first hand how it works. When police officers ran a photo of her through the app, they soon received calls from Clearview asking if they were talking to the media. Read her investigation: http://nyti.ms/36bAfTH Matt Blaze / @mattblaze : Terrific reporting by @kashhill. Today this data and tech is used to efficiently identify petty criminals, which means that next week it will be used to identify insurance risks, and the week after to locate stalking targets. Trickle down big brother. https://twitter.com/... Jameel Jaffer / @jameeljaffer : This is an amazing, scary story from @kashhill. Some of the quotes are absolutely crazy. https://twitter.com/... Jason Schreier / @jasonschreier : Even if you've grown indifferent to privacy concerns these days (or you just assume that tech companies are tracking everything we do no matter what) this is still pretty shocking. Tremendous reporting by @kashhill https://www.nytimes.com/... Kashmir Hill / @kashhill : @Aaron_Krolik Clearview says 600 law enforcement agencies are using its app. Detectives tell me it's amazing. When the founder took a photo of me while I covered my nose & mouth, it still worked, returning 7 photos of me, one 10 years old. It's insane. Read the story: https://www.nytimes.com/... Kashmir Hill / @kashhill : When @Aaron_Krolik did a forensic analysis of the app, he discovered code to pair it with augmented reality glasses, so you could theoretically identify people in real time walking down the street. Yeah, like in Terminator 2. https://twitter.com/... Kashmir Hill / @kashhill : When I first started looking into Clearview AI, it had a nonexistent office address on its website & one fake employee on LinkedIn, and no one from company would return my calls. But they knew about me and were monitoring for cops who uploaded my photo to their app. https://twitter.com/... Kashmir Hill / @kashhill : I'm not sure which is scarier/more desirable. An app that puts a name to a face in seconds, or an app that shows you all the online photos of you that you didn't realize were there. This app does both, but only law enforcement has access to it, for now. https://twitter.com/...
Context & Ripple Effects
Kashmir Hill's investigation is the moment the privacy community's long-running fear gets a name and a customer list: Clearview AI scraped billions of photos from sites like Facebook and YouTube into a search engine that over 600 law enforcement agencies — including the FBI and DHS — use to turn an unknown face into an identity. Backed by Peter Thiel, the startup even shipped code to pair with augmented reality glasses for real-time identification, and it polices its own users, contacting agencies after police uploaded a photo of Hill herself.
The story's aftermath confirms the pattern rather than correcting it: by late 2021 the CEO claimed a corpus past 10 billion images with new police tooling like deblurring and mask removal (Clearview's own growth trajectory), and a 2023 retrospective found Meta and Google deliberately withheld their face-recognition tech over privacy worries — a retreat that handed the market to startups like Clearview and PimEyes (the platform-abstention dynamic).
First-order effects
- Over 600 US law enforcement agencies, including the FBI and DHS, can now identify strangers from a single photo using a database built without those people's knowledge or consent.
- Facebook, YouTube, and other scraped platforms face immediate questions about how a startup took billions of user photos and turned them into a policing product.
Second-order effects
- Meta and Google's choice to hold back their own face-recognition capabilities leaves the identification market to unconstrained startups like Clearview and PimEyes, trading platform risk for startup risk.
- Clearview's internal monitoring — flagging agencies after a journalist's photo was uploaded — signals that usage discipline, not contracts alone, becomes the product's main governance mechanism.
Third-order effects
- If the pattern holds, every publicly posted photo functions as a biometric record searchable by the state, forcing regulators to decide whether scraping public web content counts as lawful data collection.
- The AR-glasses code points toward ambient, real-time identification as the endgame, making the current photo-search debate a staging ground for always-on surveillance policy.
The trend: Consumer web data is being converted into state surveillance infrastructure faster than privacy law can define permission boundaries, with big platforms stepping aside and startups filling the gap.