Clearview AI is quietly testing an experimental prototype called InquiryIQ, a tool designed to automate the tedious process of building digital profiles for law enforcement investigations. Discovered by WIRED in files sent to browsers before login, the tool uses AI agents to fan out across the web, analyzing images and text to uncover employers, aliases, associates, and physical characteristics of a subject. While the company insists the tool is not yet shipped to customers or planned for immediate release, its existence signals a shift from simple face recognition to comprehensive, automated intelligence gathering.
Under the Hood: Model Selection and Demographic Inputs
The prototype interface reveals that InquiryIQ was tested using models from xAI (now SpaceXAI), the company behind Grok, as well as Amazon Bedrock. The code indicates that supplying demographic data—such as age, gender, and race—helps the AI make "smarter decisions" during its search process. Clearview CEO Amos Kyler clarified that these model options were included for internal engineering comparisons rather than allowing police to choose specific AI engines. However, the reliance on xAI’s Grok is notable given the model’s history of controversial outputs, including instances of antisemitic posts and praise for Adolf Hitler following system prompt changes in 2025.
The Frictionless Investigation Problem
Privacy experts argue that tools like InquiryIQ eliminate the practical friction that once served as a check on surveillance. By compressing days of detective work into minutes, the tool lowers the cost of fishing expeditions, making it feasible to scrutinize individuals police might otherwise ignore. Boston University privacy scholar Woodrow Hartzog warns that this reduction in labor removes the natural barriers to following everyone around, potentially leading to a world where privacy protections built for a high-friction environment become obsolete.
Hallucinations and the Human in the Loop
A significant technical concern is the reliability of generative AI in investigative contexts. Michael Price of the National Association of Criminal Defense Lawyers argues that a "hallucination-prone chatbot" should not be trusted as an informant for establishing probable cause. The interface requires officers to attest that they have independently verified any automatically generated data, but critics fear the human in the loop will eventually become a rubber stamp. In a related case, United States v. Sant, a Clearview report included matches labeled "A Less Likely Result" that were accepted by an analyst, sweeping in photos of unrelated family members, highlighting the risks of over-reliance on automated matching.
Key Takeaways
- InquiryIQ is an unreleased prototype discovered in Clearview’s browser files, not a live product for police.
- The tool uses AI agents to automate web scraping and profile building, connecting scattered digital clues.
- Testing involved xAI’s Grok and Amazon Bedrock, with demographic inputs influencing search decisions.
- Experts warn that reduced investigative friction could lead to unchecked surveillance and privacy erosion.
- Human verification is required, but the risk of rubber stamping AI outputs remains a critical concern.
The Bottom Line
Clearview’s InquiryIQ represents the logical next step in surveillance tech: automating the detective’s intuition. While the prototype promises efficiency, the integration of hallucination-prone LLMs into evidence-gathering workflows creates a dangerous gap between speed and truth. Developers building for law enforcement must prioritize auditability over automation, or we risk a future where probable cause is determined by the most convincing hallucination.
Source Context
This report is based on exclusive findings by WIRED, which analyzed code and text within Clearview’s login page files. The article notes that Clearview’s database has grown to over 70 billion images since its controversial 2020 launch, and the company currently serves more than 2,000 law enforcement agencies.