Startup Clearview AI is testing an AI-based tool that automatically gathers information about a person after identifying them by their face. The prototype is called InquiryIQ, as reported by WIRED. Journalists discovered the program code in files that the Clearview AI login page transmitted to the browser even before the user logged in. These files contained interface descriptions, instructions, warnings, and a list of features. The publication did not gain access to the server-side of InquiryIQ, so WIRED could not independently confirm the full implementation of the declared capabilities. Previously, Clearview AI considered facial recognition as a starting point for investigation, and upon a match, an employee independently verified the data associated with the person. InquiryIQ automates part of the next stage: after entering initial data, the system should search for information on the internet and through images, open pages, and analyze the found photographs.
Based on the results, the prototype builds a Candidate Graph, which is a graph of possible identities and connections between them. The profile may include presumed addresses, phone numbers, employers, pseudonyms, social media pages, arrest records, and related persons. The interface provides for specifying age, gender, and race, and according to the studied code, these parameters should help the AI make more informed decisions during the search. How exactly demographic characteristics affect the results remains unknown. The company claims that law enforcement agencies have never used InquiryIQ, the prototype has not been offered to clients, and its release in its current form is not planned. The startup warned in the interface that automatically found information may be incorrect, and the user must accept or reject each finding and confirm independent verification. Boston University law professor Woodrow Hartzog doubts this is sufficient, calling human oversight a weak consolation, as over time, the reviewer risks becoming a formal confirmer of the system's decisions.
In the InquiryIQ interface, it was possible to choose a system for analysis. Among the options, WIRED found a model from SpaceXAI and Amazon Bedrock, a platform for accessing third-party developer tools, while Amazon representatives told the publication that AWS is not involved in creating the tool. Experts interviewed noted that such systems can reduce manual work that previously took investigators hours or days. Hartzog pointed out the downside of automation: the laboriousness of collecting disparate information itself limited the scale of surveillance, and InquiryIQ potentially lowers this barrier. George Washington University professor Andrew Guthrie Ferguson noted that preserving prompts, search queries, and action sequences will simplify the audit of investigations, calling it the bright side of a potentially problematic change. Clearview AI claims a database of more than 70 billion images and works with over 2000 US law enforcement agencies, and in 2023, French authorities fined the company 20 million euros for illegal processing of personal data.