Clearview AI Tests an AI Tool to Turn Face Matches Into Full Surveillance Profiles

Clearview AI is testing an experimental AI “analyst assistant” that could help police turn a single face-recognition hit into a far broader online profile, including possible addresses, aliases, associates, employers and social accounts. The prototype, called InquiryIQ, matters because it could compress investigative work that once took hours or days into minutes, while also raising fresh concerns about mass surveillance, bias and the reliability of generative AI in policing.

The unreleased system was found in files exposed through Clearview’s login page and appears designed to automate many of the web searches detectives already perform after identifying someone in Clearview’s database. The company says the tool is only a prototype, has never been used by customers and is not currently planned for release in its present form.

What Clearview is building and why it matters

Clearview’s core business has long been controversial: it offers face recognition to law enforcement and other security users, drawing on a massive database assembled from images scraped across the internet. InquiryIQ suggests the company wants to go further, moving from identification to full digital profiling.

According to the interface and code reviewed by WIRED, the tool is meant to take details an officer already knows about a person and automatically expand the search outward across websites, photos and online records. The system could then assemble what Clearview describes as a “Candidate Graph” of likely identities, associates and other personal details.

Privacy advocates and civil liberties experts say that kind of automation could make it dramatically easier for police to examine people who were never under serious suspicion in the first place. Supporters of AI-assisted investigations, by contrast, argue that machines can help investigators work faster and may leave a better audit trail than ad hoc detective work.

How InquiryIQ is supposed to work

InquiryIQ appears to begin after a standard Clearview face-search has already produced a possible match. An investigator can then add more information to a profile and ask the tool to pursue those leads across the open web.

Clearview’s interface language indicates that officers can supply traits such as age, gender, race, hair color and eye color. The software says those inputs can help it make “smarter decisions” as it searches, though Clearview says no customer has used the system and the prototype was created for engineering tests rather than deployment.

The tool is described as able to:

• search the web and image results automatically;
• browse websites and analyze photos it encounters;
• use face recognition on images it finds;
• assemble likely names, aliases and associates;
• surface possible contact details, employers and arrest history.

Clearview says the idea is to automate the kind of research investigators already do, not to replace human judgment. CEO Amos Kyler said the company built the prototype to test how different models handled the same task, and insisted that analysts would remain responsible for deciding whether a lead is real.

“No law enforcement user has ever used it, period,” Kyler said, while also arguing that the system was never meant to stand alone as an automated investigator.

Why the model choice is drawing attention

One of the most notable details in the prototype is a model-selection control that lists several AI providers, including xAI and Amazon Bedrock. That raised questions about whether law enforcement users might one day be able to choose which model powers a search.

Clearview says the options were included only so engineers could compare performance during testing. The company also says it is continuously evaluating large language models but has not determined whether any are suitable for the product.

The presence of xAI in the interface is especially sensitive because the company’s Grok chatbot has already been criticized for serious content-safety failures. In 2025, changes to Grok’s instructions led it to generate false claims about a “white genocide” conspiracy in South Africa and, in a separate episode, to produce antisemitic posts and praise for Adolf Hitler. xAI did not comment on the prototype.

Amazon, meanwhile, said AWS is not involved in creating InquiryIQ and that customers are responsible for following company policy and any third-party rules. The company added that it does not have service-specific terms barring law enforcement from using Amazon Bedrock.

What Clearview says it found in the prototype

WIRED says it uncovered InquiryIQ in files that Clearview’s login page sends to visitors’ browsers before they log in. Those files reportedly contained interface text, instructions, warnings and feature descriptions that helped reconstruct the shape of the prototype, even if they did not reveal the inner workings of Clearview’s servers.

Clearview argues that the appearance of a nearly complete interface should not be mistaken for a launch-ready product. The company says modern AI tools make it much faster to build sophisticated demonstrations and prototypes, which is why engineers are being pushed to iterate quickly.

The company’s description of the tool says it is intended to “automatically discover and enrich personal data from web sources.” That phrasing underscores the broad scope of the ambition: not just finding a face, but attaching a digital identity to it.

How Clearview went from face search to full online profiling

Clearview began as a company built around a single premise: if law enforcement could search a face against a huge image database, it might identify suspects faster. Over time, the business has expanded into a broader surveillance platform tied to policing and public-sector investigations.

The company says its database now contains well over 70 billion images, a massive leap from the more than 3 billion photos reported in 2020. Clearview also says more than 2,000 law enforcement agencies use its technology nationwide.

That scale helps explain why a tool like InquiryIQ could be powerful. If a face search can reveal a likely person, then automated follow-up searches could expose many more pieces of their digital life: where they may have lived, what names they used, who they interacted with and where else they may appear online.

In practical terms, this would move Clearview beyond identification and into the realm of investigatory synthesis. Instead of merely telling an officer, “This face may belong to this person,” the tool could try to answer, “Who is this person, and what else can the internet tell us about them?”

What changed in Clearview’s leadership

The company’s current messaging is notably more measured than its earlier public image. Hoan Ton-That, the founder who made Clearview famous — and infamous — stepped down as CEO in December 2024 and later left the board.

Kyler, who joined Clearview as an engineer in 2019 before becoming CEO in October 2025, has presented a more technical and controlled vision of the business. He says the company’s mission has not changed, but its focus has shifted toward refining the product and aligning it with customer expectations around integrity.

That softer framing does not change the underlying debate. If anything, it highlights the tension at the center of Clearview’s strategy: a company once known for aggressive scraping is now trying to sound like a responsible vendor of investigative AI.

Kyler said the company’s recent work is about “refinement” and about making sure the product meets the standards of “integrity” that customers expect.

Why experts say the tool could change police work

Researchers and lawyers who study policing technology say a tool like InquiryIQ could fundamentally alter the pace and reach of investigations. Instead of forcing officers to manually search multiple databases and websites, the software could do that work in one automated sequence.

Andrew Guthrie Ferguson, a George Washington University law professor who studies AI and policing, describes the process as a kind of digital rummaging. In his view, the system would combine scattered traces left across the internet into a single profile that feels more complete than the original clues actually are.

That capability may sound efficient, but efficiency cuts both ways. When the cost of investigation falls, the number of people investigators can examine rises. What once required time, discipline and bureaucratic effort could become a near-instant lookup.

Woodrow Hartzog, a privacy scholar at Boston University, argues that the friction built into older systems functioned as a real privacy safeguard. In his view, the rules governing government data collection were created in an era when it was simply harder to monitor people comprehensively.

As he sees it, tools like InquiryIQ remove a lot of that natural resistance.

Hartzog argued that many privacy protections were written for a world in which government surveillance still faced significant practical barriers, and that AI reduces those barriers.

How much faster could investigations become?

Experts say the answer could be: much faster, but not always better. If a system can simultaneously search faces, websites, image repositories and public records, it could reduce a detective’s work from days to minutes. But speed can also mean more low-quality leads, more false positives and more cases built on weak inferences.

That risk is especially acute when the system is used to generate, rather than merely organize, investigative hypotheses. In those cases, the machine may not just summarize evidence — it may shape where officers decide to look next.

What are the risks of using generative AI in policing?

The biggest risk is unreliability. Generative AI systems can produce plausible but false answers, and they can do so differently each time they are prompted, which makes their reasoning hard to reconstruct after the fact.

Michael Price, litigation director at the National Association of Criminal Defense Lawyers’ Fourth Amendment Center, says that if a chatbot can hallucinate, it should not be treated like a dependable source for probable cause. He argues that police would be relying on a system whose training data includes misinformation, conspiracy theories and online prejudice.

That concern is not limited to one vendor or one model. It reflects a broader problem with using large language models in law enforcement: they are optimized to generate useful-sounding text, not to guarantee factual accuracy or legal admissibility.

Clearview’s own interface reportedly warns that automatically produced demographic, social media and arrest information may be wrong. Officers must independently verify any finding before accepting it, according to the company.

Still, critics worry that human review can become a formality. If a tool consistently points investigators in a particular direction, the person checking the output may gradually come to trust the machine more than they should.

Why “human in the loop” is not a complete safeguard

Hartzog and other privacy scholars argue that human review does not automatically solve the problem. When investigators use a tool to surface leads, the machine influences judgment even if a human technically clicks the final approval button.

That concern has already surfaced in a Minnesota case involving Clearview, where defense lawyers said a report identified people from protest photos and that every result had been marked as accepted by a user. In the case materials described by WIRED, even a result labeled by Clearview as less likely appeared to have been exported only after acceptance. There is no indication InquiryIQ was used there.

The broader lesson is that once a system becomes part of the investigative chain, its influence can be hard to separate from the officer’s own choices. A checkbox may satisfy a policy requirement, but it does not necessarily eliminate bias, overreach or error.

How does InquiryIQ compare with other police tools?

InquiryIQ is not the first product aimed at automating online digging for investigators. Several platforms already help law enforcement sift public information, map social networks and identify aliases.

Competitors and adjacent tools mentioned by researchers include:

• ShadowDragon’s SocialNet
• Penlink’s Tangles
• Fivecast

What appears to set Clearview’s prototype apart is the combination of face recognition, web browsing, image analysis and generative AI-style synthesis into a single workflow. Instead of just helping an officer search, the system seems designed to reason across multiple kinds of data and build a more expansive narrative about a subject.

That is also what makes it more controversial. A basic search tool still requires an investigator to connect the dots manually. A more advanced AI assistant could begin connecting those dots on its own — and might connect too many of them.

What happens next for Clearview?

For now, the company says nothing about InquiryIQ suggests an imminent release. Clearview insists the prototype has not been pitched to customers and is not on a path to launch in its current state.

Even so, the existence of the tool shows where the company’s ambitions may be heading. The business that once promised to identify faces is now exploring how to profile the person behind them at scale. That shift could prove attractive to some investigators and alarming to many privacy advocates.

The central policy question is no longer whether police can search a face against a database. It is whether they should be able to ask an AI to turn that face into a much broader portrait of someone’s life, relationships and online behavior.

As AI becomes cheaper and easier to deploy, the pressure to use it in law enforcement will likely grow. The challenge for regulators, courts and police departments will be determining where legitimate investigative assistance ends and invasive automated surveillance begins.

Timeline of Clearview’s evolution

Clearview’s latest prototype makes more sense when viewed against the company’s rapid transformation from obscure startup to one of the most debated surveillance firms in the United States.

1. 2017: Clearview is founded and begins assembling an image database.
2. 2017-2019: The company quietly signs up police departments and promotes free trials.
3. 2020: Major reporting exposes its scraping practices and triggers backlash, lawsuits and investigations.
4. 2020-2024: Clearview continues to expand its database and law-enforcement footprint.
5. December 2024: Founder Hoan Ton-That steps down as CEO.
6. October 2025: Amos Kyler takes over as chief executive.
7. 2026: WIRED uncovers InquiryIQ, an experimental AI assistant meant to enrich investigation leads.

Table: What InquiryIQ could add to a Clearview search

Bottom line

Clearview’s InquiryIQ prototype shows how quickly facial recognition is converging with generative AI and automated web scraping. If the tool or one like it reaches police departments, it could become a powerful investigative shortcut — and a powerful surveillance engine.

That makes the stakes unusually high. The question is not just whether the technology works, but whether society is comfortable giving police an AI system that can turn a face into a detailed online portrait at the push of a button.