Clearview AI Trials AI Tool to Help Police Discover Your Online Activities

Clearview isn’t the first to automate the search of publicly accessible information. Other intelligence platforms aimed at law enforcement, such as ShadowDragon’s SocialNet, Penlink’s Tangles, and Fivecast, assist investigators in revealing aliases and connections while mapping an individual’s digital presence.

Andrew Guthrie Ferguson, a law professor at George Washington University focusing on AI and policing, refers to this type of automated investigation as “digital rummaging.” He explains, “They’re essentially creating a profile based on all the random digital traces you’ve left online.” In effect, this involves collecting disparate pieces of information that were once isolated and hard to link.

Woodrow Hartzog, a privacy expert from Boston University, contends that the effort involved in investigative work acted as a practical safeguard against surveillance. By drastically decreasing the labor associated with probing into a person, tools like InquiryIQ also forgo these safeguards. “The privacy protections we currently have were largely established in a context that assumed a certain level of difficulty in government data collection about individuals,” he states. “Many regulations never existed simply because they weren’t necessary due to practical barriers preventing constant monitoring.”

After an InquiryIQ search concludes, the interface is designed to show the officer the identified individuals, connections, and other relevant details that the system has discovered, which the officer can approve or dismiss before they are incorporated into the profile. Clearview cautions that the automatically generated demographic, social media, and arrest information “may not be accurate.” Officers must confirm that they have independently verified any findings before acceptance.

Kyler emphasizes that human oversight is integral to Clearview’s framework. The system aims to highlight potential leads, not to ascertain truths, he states. “That’s the analyst’s role—to assess what is true, what isn’t; what’s noise and what’s reality.”

“Having a human involved offers a bit of cold comfort,” Hartzog remarks. Over time, investigators might rely on automated systems to the extent that the individual reviewing the results becomes merely “a sort of rubber stamp.”

For instance, in United States v. Sant, a Minnesota case involving undercover agents from Homeland Security Investigations monitoring political activists, defense attorneys obtained a Clearview report based on around 15 years of protest images. Every matched result was marked “Accepted by Guy Gino,” including one that Clearview itself categorized as “A Less Likely Result”—which, as noted in the report’s footnote, can only be exported if an operator accepts it. Defense counsel claims the report included images of an entirely different individual along with his pregnant wife and young daughter. (There is no evidence that InquiryIQ was employed in this case.)

Pitfalls and Silver Linings

The prototype reviewed by WIRED displayed varying results from InquiryIQ based on the chosen model. Its interface features a control for selecting the model that conducts the research, listing options like xAI and Amazon Bedrock, a platform for accessing models created by others.

In 2025, xAI reported an unauthorized alteration to Grok’s system prompt that led the chatbot to incorporate claims about a supposed “white genocide” in South Africa during unrelated discussions. Less than a month later, after another adjustment to its instructions, Grok generated antisemitic comments and admiration for Adolf Hitler. SpaceXAI did not respond to a request for comment.