Clearview AI Built an AI Prototype for Police Profile Searches

Clearview AI has built and tested InquiryIQ, an unreleased AI prototype intended to turn a facial-recognition result into a broader police research profile. The system is designed to search the web and images for clues about a person’s possible identity, associates and online footprint.

Clearview says InquiryIQ has never been used by a law-enforcement customer. The company also says it has not pitched or shipped the tool and does not currently plan to release it in its present form. The development is therefore a view into an internal prototype, not a deployed police product.

From a face-search lead to a web profile

The intended workflow begins after an investigator runs a Clearview face search and adds details they consider relevant. InquiryIQ is designed to conduct web and image searches, browse webpages, and apply facial recognition to photographs it finds. Its interface says age, gender and race can help the system make smarter decisions during searches, though Clearview declined to speculate on how those inputs might shape output.

What the prototype is designed to return

• A Candidate Graph of possible identities and associates.
• Possible addresses, phone numbers, employers, social-media accounts, arrest history and aliases.
• Findings for an investigator to accept or reject before adding them to a profile.

Verification remains the stated boundary

Clearview’s interface warns that automatically generated demographic, social-media and arrest information may not be accurate. Before accepting a finding into a profile, an officer must attest that they independently verified it. Clearview says the system is meant to generate leads for human analysts, who remain responsible for judging what is true.

That human-review requirement does not settle the concern for Woodrow Hartzog, a Boston University privacy scholar. He told WIRED that reviewers can become “a sort of rubber stamp” as investigators defer to automated systems, and argued that the labor of gathering scattered information has historically acted as a practical limit on surveillance.

What the prototype evidence can show

WIRED found InquiryIQ interface code and descriptive text in files sent to visitors’ browsers before sign-in. The material shows how Clearview has designed and described the tool, but does not establish its server-side behavior or how well it works in an investigation.

The interface included xAI and Amazon Bedrock among model options tested by Clearview engineers. Clearview says those options were for internal performance comparisons, not for law-enforcement users to choose during an investigation, and says it has not assessed any model’s suitability for InquiryIQ.