B ∙ R ∙ A I ∙ N

Federal use of facial recognition and related surveillance technologies has expanded decisively into the monitoring of domestic protest activity, according to a growing body of reporting and firsthand accounts from Minneapolis, where Immigration and Customs Enforcement (ICE) and Federal Bureau of Investigation (FBI) agents have deployed an unusually dense mix of biometric, social media, and data-analytics tools during large-scale enforcement operations.

The picture that emerges though is not of a single technology used in isolation, but of an integrated surveillance stack that allows federal agents to identify, locate, and track individuals who appear near enforcement actions or protests, even when those individuals are U.S. citizens engaged in lawful activity.

Facial recognition is a central component, but it is only one layer in a much broader system.

One incident illustrates how directly these tools are now intersecting with everyday encounters. On January 10, Nicole Cleland, a 56-year-old Richfield, Minnesota resident who volunteers with a local watchdog group that observes ICE, was driving behind an ICE agent when the agent abruptly turned down a series of one-way streets, stopped, and approached her vehicle.

According to Cleland, the agent addressed her by name even though the two had never met. He told her that he had facial recognition technology and that his body camera was recording.

Cleland was not alone. Another woman peacefully protesting recorded her encounter with an ICE agent who photographed her and told her she was now in a “domestic terrorism” database.

Local activists say at least seven American citizens were told by ICE agents in and around Minneapolis that they were being recorded with facial recognition technology.

Videos posted to social media show agents informing people that their faces would be added to a database. None of those individuals had given consent.

These encounters underscore how facial recognition is being used not only to identify undocumented immigrants, but also to track protesters, observers, and critics of ICE operations.

Current and former Department of Homeland Security (DHS) officials have said that agents in Minneapolis are using at least two facial recognition systems. One is produced by Clearview AI, a controversial biometrics firm known for scraping billions of images from the open internet. Its database held more than 50 billion images in June 2024.

The other is a newer mobile tool known as Mobile Fortify, which allows ICE agents to scan faces in the field using cellphones and route images into backend identification systems.

Photographs and videos from Minneapolis show agents using Mobile Fortify during protests, scanning faces with phones while wearing tactical gear. Mobile Fortify, originally developed for Customs and Border Protection but later adapted for broader DHS use, relies on biometric technology supplied by NEC Corporation.

A DHS spokeswoman has described Mobile Fortify as a lawful tool used nationwide, but the reporting shows it is now being applied near protest activity.

Clearview AI’s role adds another layer of concern. DHS acknowledged in a 2024 report that it used Clearview’s technology for child exploitation investigations. However, a $3.75 million contract awarded to the company in September expanded its use to investigations involving attacks on law enforcement.

Critics argue that this broader authorization creates a pathway for protest related facial searches, even when individuals are not suspected of violent crimes.

Clearview has said its work with DHS is focused on child exploitation and cybercrime, but the Minneapolis accounts suggest a wider operational footprint.

Facial recognition does not operate alone. ICE is also drawing on an array of digital surveillance tools that extend identification into real-time tracking and dossier building.

Officials say agents are tapping into a database built by Palantir Technologies that combines government records with commercially sourced data to help locate individuals.

In April, DHS awarded Palantir a nearly $30 million contract to build an AI-backed system designed to identify and track people for deportation, with a prototype due last fall. According to current officials, the system is already being used in operations.

Palantir has said that its work with ICE is intended to identify noncitizens, but civil liberties advocates warn that when such systems integrate multiple data sources, the boundary between targeting noncitizens and monitoring citizens can become porous, particularly when protest footage and social media activity are used as inputs.

ICE has also significantly expanded its social media and cellphone monitoring capabilities. Procurement records show that DHS spent nearly $10 million acquiring multiple tools that allow agents to scrape social media, collect data from brokers, and in some cases access the content or location data of mobile phones.

Among those vendors are Paragon Solutions, an Israeli-founded, U.S.-owned surveillance technology firm whose technology can be used to remotely access phones, and Penlink, which provides software to aggregate social media data and build detailed profiles tied to online accounts.

Together, these tools allow agents to move from a face in a crowd to a name, from a name to online activity, and from online activity to real-world location data. A partial facial recognition match can be strengthened using social media posts, commercially purchased datasets, or prior encounters stored in government systems.

The result is an identification process that can reach high confidence even when no single data point is definitive on its own.

This architecture is reinforced by interagency cooperation. The FBI has its own facial recognition capabilities through the Next Generation Identification system, which can search hundreds of millions of images sourced from federal and state partners.

In at least one Minneapolis-related case, FBI agents used facial recognition on still images extracted from publicly posted protest videos, generating an identification that later appeared in a federal criminal complaint related to alleged property damage.

The case showed how protest footage posted by private citizens can be ingested into federal biometric systems without warrants or notice to the individuals depicted.

Oversight bodies have warned for years about the scale of these databases and the limited safeguards governing their use. Federal policy typically characterizes facial recognition results as “investigative leads,” but civil liberties experts argue that this language understates the consequences of being identified.

Once a name is generated, it can trigger surveillance, follow-on investigation, and long-term retention in law enforcement systems, even if no arrest is made.

The Minneapolis operation highlights how these capabilities intersect with constitutionally protected activity. Protesters, legal observers, journalists, and bystanders often appear in the same footage.

Identification can occur preemptively, without suspicion of criminal wrongdoing, based solely on proximity to enforcement actions or participation in demonstrations.

The expansion of ICE’s technology has been fueled by a dramatic increase in funding. Last July, President Trump signed legislation increasing ICE’s annual budget to roughly $28 billion, up from about $8 billion, making it the most heavily funded law enforcement agency in the federal government.

Since then, the agency has accelerated procurement of biometric, analytics, and surveillance tools.

ICE recently issued a request for information seeking new ways to integrate “big data” and advertising technology into its operations, asking vendors what data they could provide on people, devices, locations, and transactions, and whether individuals could be searched by identifiers such as phone numbers or accounts.

Lawmakers have begun to push back. In November, Democratic senators asked ICE to suspend its use of facial recognition in U.S. cities and provide more information about how the technology is being deployed.

Advocacy groups including the American Civil Liberties Union have sued DHS over the Minneapolis operation, arguing that the conglomeration of technologies gives the government unprecedented power to monitor and deter dissent.

For individuals caught in the system, the consequences can be immediate and personal. Three days after her encounter with the ICE agent, Cleland received an email informing her that her Global Entry and Transportation Security Administration PreCheck privileges had been revoked, without explanation.

She has since joined a lawsuit challenging ICE’s treatment of observers, asking why her travel status was stripped after she was identified during protest monitoring.

“I am a totally average American,” Cleland said in a declaration filed with the court. “I cannot abide by what is happening right now.”

For the biometrics industry, Minneapolis offers a stark case study in how facial recognition systems migrate across use cases once deployed at scale.

Technologies procured for border security, child exploitation investigations, or serious crime are now embedded in domestic protest policing, augmented by data brokers, social media analytics, and AI-driven tracking tools.

The capability is mature, the data streams are plentiful, and institutional incentives favor expansion rather than restraint.

What remains unsettled is whether oversight and legal frameworks will evolve to match this reality. The Minneapolis operation has made visible a surveillance pipeline that has existed for years but largely operated out of public view.

As biometric identification becomes increasingly frictionless, the central question is no longer whether federal agencies can identify protesters, but what limits, transparency requirements, and accountability mechanisms will govern how that power is used.