Clearview AI, the facial recognition company already known for turning a photograph into possible names and online images, has been quietly testing a more ambitious prototype. According to reporting published Thursday by WIRED, the experimental system, called InquiryIQ, is designed to start with details from a facial recognition search and then hunt across the open web for a much larger picture of a person’s life. The distinction matters.
Clearview’s existing product compares a photo against a vast collection of publicly available images and returns possible matches, along with links to the pages where those photos appeared. InquiryIQ, as described in interface code reviewed by WIRED and in comments from the company, would take those details further. It is built to open webpages, analyze images it finds, run additional face comparisons, and assemble possible employers, aliases, associates, addresses, phone numbers, social media accounts, arrest history, and physical characteristics into a single investigative profile.
Clearview has been unusually clear about one point: InquiryIQ is not a finished product in the hands of police. The company told WIRED the tool is a prototype that has never been pitched or shipped to customers and is not currently planned for release in its present form. Chief executive Amos Kyler said no law enforcement user has ever used it. That claim should be taken as the company’s position, not as proof of how the technology would perform if it were later deployed.
There is no verified public record showing that any police department or federal agency has tested, demonstrated, or been given access to InquiryIQ. WIRED found the prototype in files that Clearview’s login page sends to a visitor’s browser before anyone signs in. Those files contain interface text, warnings, and feature descriptions. They do not reveal what happens on Clearview’s servers, how accurate the system is, or whether it works as advertised.
According to that text, InquiryIQ is meant to “automatically discover and enrich personal data from web sources.” As it searches, it is designed to build what Clearview calls a “Candidate Graph” of possible identities and associates. The interface also asks for personal attributes. Supplying age, gender, and race, it says, can help the system make “smarter decisions” as it searches.
Hair color and eye color can be added as well. WIRED reported that one of the models Clearview tested to power those decisions came from SpaceXAI, the Elon Musk company behind the Grok chatbot, after SpaceX and xAI merged earlier this year. The prototype interface also listed Amazon Bedrock, a platform for accessing other companies’ models. Kyler said those options were included so engineers could compare models during testing, not so officers could pick which model ran a search.
Clearview said it is still evaluating large language models and has not assessed the suitability of any of them for InquiryIQ. Amazon said AWS is not involved in developing the tool. That is what makes InquiryIQ different from ordinary facial recognition. A conventional Clearview search can connect a face to other photos and, often, to a name or an online page.
InquiryIQ is designed to treat that first hit as a starting point, then fan out through public web material the way an analyst might, only faster. In theory, a single photo could lead to a cluster of possible identities, friends or coworkers, job listings, old social media profiles, and other scattered clues that once lived in separate corners of the internet. Clearview disputes the idea that this would be an automated investigator.
Kyler described it as a limited way to automate web searches detectives already perform. In his account, the system would take a “factoid,” search around it, and ask whether the result is related. When a search finished, an officer would see proposed identities and connections, accept or reject them, and attest that each accepted finding had been independently verified. The interface warns that automatically generated demographic, social media, and arrest data “may or may not be accurate.”
Privacy scholars say the risk is not only a wrong name. It is the collapse of time and effort that once made broad digital searches expensive. Andrew Guthrie Ferguson, a George Washington University law professor, called this kind of work “digital rummaging,” a profile built from random clues a person left online. Woodrow Hartzog, a Boston University privacy scholar, argued that older privacy rules assumed friction.
If machines can follow those trails in minutes, he said, investigators may later treat human review as a rubber stamp. Clearview already carries a long record of controversy. In 2020, reporting revealed that it had scraped billions of photos from social media and other websites without people’s consent. Lawsuits and regulatory actions followed.
A 2022 settlement with the ACLU largely confined U.S. sales of its face database to government agencies. The company now says its image collection exceeds 70 billion photos and that more than 2,000 law enforcement agencies nationwide use its technology. Those figures describe Clearview’s existing facial recognition service, not InquiryIQ. Other firms already sell open source intelligence tools that map digital footprints.
What is new here is the proposed chain: a face, then a name, then an automated sweep of the public web. Generative models can also return different answers from the same starting point, which critics say could make it hard to reconstruct why one lead was chased and another ignored. Michael Price of the National Association of Criminal Defense Lawyers’ Fourth Amendment Center warned that unreliable chatbot output is a weak foundation for investigative decisions.
If InquiryIQ stayed on the lab bench, the debate would be narrower. If a later version reached police, the question would be larger: how much of a person’s public digital life should a face be allowed to unlock, and how quickly.
