PimEyes

PimEyes is the most powerful publicly available facial recognition search engine, and that fact alone is the story. For journalism, it can be useful — verifying whether a person in one photo is the same as in another, identifying anonymous figures in crowd photos, tracing image reuse across the open web, finding additional photos of a public figure during background research. The Bellingcat-style OSINT community uses it routinely for identification work. The New York Times called it "alarmingly accurate." The BBC called it "facial recognition on steroids." Both descriptions are correct. But this is not a tool that can be recommended without a long ethical discussion. PimEyes scrapes billions of face images from the open web — news sites, blogs, personal pages, court records, conference photos, university directories, anything indexable that contains a human face. It does not scrape Facebook, Instagram, or video platforms (officially). It builds biometric vectors from those faces and lets anyone with $30 search them. The opt-out process exists, but multiple investigations have shown opt-out is unreliable, the company is structurally opaque, and the tool has been used to identify children, stalk private individuals, and dox protesters and federal employees. The owner is Giorgi Gobronidze, a Tbilisi-based AI academic who bought PimEyes in December 2021 through a Dubai-registered shell corporation, EMEARobotics. The company's prior owners were a Polish team, then a Seychelles shell. Legal entities tied to PimEyes include corporations in Belize, Poland, Dubai, and the United States. An Illinois Biometric Information Privacy Act (BIPA) plaintiff's attorney spent two years trying to serve the company notice of a lawsuit and could not find a valid contact in any of those jurisdictions. This is not how a trustworthy data processor structures itself. The misuse record is severe. The New York Times, NPR, Washington Post, and Business and Human Rights Resource Centre have documented PimEyes returning images of children labeled "potentially explicit." In October 2023, after sustained pressure, PimEyes added age detection to block child searches. Harvard students built a Ray-Ban Meta glasses demo using PimEyes to identify strangers in real time and pull their addresses and phone numbers. Activists have used it to dox ICE officers. UK and German privacy regulators have opened investigations. An Illinois BIPA class action remains ongoing. For journalism, the honest framing is this: PimEyes is genuinely useful for legitimate identification work, especially in human-rights investigations where you have a perpetrator photo and need to find the person's name. It is also a tool that, by existing, makes mass surveillance of ordinary people trivially cheap. Using it for journalism funds a company whose business model depends on private individuals being identifiable. There is no clean answer. If you use it, use it sparingly, only for clear public-interest identification, never on private individuals or minors, and never as the only basis for identification — always corroborate with other evidence. Consider whether your story can be done with FaceCheck.id, Yandex reverse image search, or Google Lens instead, none of which are clean either but each of which makes different tradeoffs. Document your methodology in the story. If you used facial recognition to identify someone, say so, and explain how you verified the match. Readers should know.