ICE technologies and the perverse automated system for the persecution of Latino migrants

The Trump administration has turned U.S. Immigration and Customs Enforcement (ICE) into a mass surveillance agency. With a 2025 budget exceeding $28 billion (three times that of the previous year), the agency has built a network that links government databases, purchases data from private companies, and deploys technology on the streets.

A Mexican migrant found this out when ICE tracked the remittances he was sending to his family in Mexico, identified his address in Hawaii, and arrested him, even though he had no criminal record. In Chicago, a Colombian woman was detained in what appeared to be a routine traffic stop: her vehicle had been flagged in advance by cross-referencing state records with federal databases. Another Mexican migrant, who sold flowers in a city near Los Angeles, was detained for weeks because ICE mistook him for a suspect with a similar name, despite having documents proving his innocence. Three cases, three different technologies, the same system.

The agency has spent years building this surveillance system, a data architecture originally developed after 9/11 to “fight terrorists,” now redirected toward migrant communities that work and sustain the country’s economy. A funnel that begins with data generated by the people themselves (remittances, vehicle registrations, cell phone locations), goes through cross-referencing with government and commercial databases, and ends in a street operation where an ICE agent, holding a phone, decides someone’s fate in seconds. A system that, in this process, does not distinguish between those who have a deportation order and those who simply exist in its databases.

Stage 1: The machinery of mass data collection

The system’s starting point is data. ICE does not collect most of this information itself; it buys it. Every time someone pays an electricity bill, sends a remittance, obtains a driver’s license, or posts on social media, that information can end up in the agency’s hands. Companies like Thomson Reuters and LexisNexis gather data from multiple sources on a single person—credit histories, addresses, phone numbers, vehicle records, family ties—and sell it to ICE. No court order is required, since the law does not mandate one when the data is sold by a private company, and this is precisely the gateway through which ICE gains access.

Cell phones also give people away. Apps that millions of people use every day constantly track their location, and companies like Venntel and PenLink capture and resell those GPS movement records to the government. Out on the streets, the Flock Safety network operates 100,000 cameras across 49 states, scanning 20 billion vehicle license plates every month.

Biometric data adds another layer. Driver’s license photos are stored in state databases that ICE can access. Clearview AI went even further: it scraped images from social media and news sites to build a database of more than 50 billion faces, and ICE signed a contract to use this database. Finally, information already held by the state is also available, including tax returns, Medicaid medical records, and Social Security data. In 2025, a federal judge ruled that the Internal Revenue Service (IRS), the U.S. tax authority, had violated the law approximately 42,000 times by sharing taxpayer data with ICE.

Stage 2: Artificial intelligence turns data into targets

The information collected in the first stage is not enough. The system needs to turn it into traceable individuals, and that’s where companies like Palantir come into play. Founded with CIA funding, this tech company has held contracts with ICE since 2011. Its platform cross-references all information sources from the previous stage and builds profiles with probable locations. This is not a manual search; it is an AI model that processes thousands of variables to determine where someone might be at any given moment.

The most revealing tool Palantir has developed for ICE is ELITE. According to an agent who testified under oath in court, it works like a “Google Maps for people subject to deportation.” Essentially, it displays the probable location of each target on a map, along with a probability score. To calculate this, it cross-references, for example, the address registered with Medicaid with the most recent GPS record and the last time their license plate was captured by a camera. In addition, it builds an automated dossier based on criminal records and court orders. The system has internal restrictions on what it can do and who it can flag as a target. However, in certain operations that the agency refers to as “special,” different filters are applied to identify individuals who, under normal circumstances, would not qualify as targets. So far, ICE has not publicly explained when this occurs or who authorizes it.

In 2025, the immigration agency paid an additional $30 million to expand this system and build ImmigrationOS, a platform designed to track individuals from the moment they enter the country until they are deported, with real-time monitoring. A person who enters the system as data ends up reduced to a case file with a status and an estimated departure date, without ever having been heard or given the opportunity to challenge the algorithm’s decision about them.

Stage 3: Any street can be a surveillance zone

Once a target is identified, the system moves from algorithms to the streets. The agent arrives with an app called Mobile Fortify on their government-issued phone, points it at anyone’s face (not necessarily the one they were looking for), and, within seconds, matches that image against millions of photos collected in Stage 1. It is worth noting that these photos were taken without proper notice: people do not know that one day these records will become part of a mass identification system. If there is a match, the system returns the person’s name, date of birth, nationality, and whether a deportation order exists. If the facial image is not sufficient, another BI2 Technologies tool repeats the process using iris recognition. All of this happens on the street, without the person having done anything wrong, without their knowledge, and without any judicial authorization.

The most concerning aspect of Mobile Fortify is that it was deployed while still in the testing phase, without any privacy impact assessment. Worse still, it lacks accuracy. In a documented case before an Oregon court, the system scanned the same woman twice and returned two different names. Nevertheless, ICE treats its results as definitive proof, even over a birth certificate.

There are also tools that operate even more covertly. Cellular simulators, also known as IMSI Catchers, disguise themselves as cell towers to force nearby phones to connect and identify which devices are in a given area. The immigration agency has purchased vehicles equipped with these devices to deploy them covertly in specific neighborhoods. For those going through the immigration process, there is SmartLINK, an app that ICE requires individuals to install on their personal phones, which tracks their location and requires periodic facial recognition verification. A Colombian migrant in California discovered this when the app blocked her from traveling to another state. When she asked why, an officer told her she had done nothing wrong; it was a discretionary decision by the agency.

When migrants are detained, they experience an even greater loss of privacy. Devices from companies like Cellebrite and Magnet Forensics can unlock and extract the full contents of a phone within minutes, including messages, photos, location history, and contacts. In fact, encrypted apps like WhatsApp cannot protect this information under these circumstances. Encryption prevents messages from being intercepted as they travel over the internet, but once a phone is unlocked in an agent’s hands, that encryption becomes irrelevant. A detained person’s conversations can serve as a starting point for tracking others who have never had any contact with ICE.

Protecting yourself is also a form of resistance

Understanding how this system works is already a form of resistance. For those who are migrating or planning to do so, this knowledge can make all the difference from day one. Your cell phone is your most vulnerable source of data. Using Signal instead of WhatsApp significantly reduces the data trail: Signal does not store metadata or share information with third parties. Reviewing which apps have access to your location and revoking unnecessary permissions limits what companies can collect and sell. In areas with a high ICE presence, a Faraday bag blocks all cell phone signals, preventing cellular simulators from detecting the device. And on social media, posting locations or tagging places means voluntarily giving away information that the system would otherwise pay to obtain.

However, individual measures have their limits; the system described cannot be dismantled with an app. The Electronic Frontier Foundation publishes free, up-to-date guides on protecting digital privacy in high-risk situations; its Surveillance Self-Defense section is a practical starting point. Organizations such as the ACLU, the National Immigration Law Center, and Just Futures Law are actively litigating on this front. Sharing their work and demanding that elected representatives take a stand are also ways to take action.

Does people’s freedom depend on an algorithm?

What Edward Snowden revealed in 2013 about the NSA’s mass surveillance seemed, at the time, to be the limit of what was imaginable. Today, ICE’s actions go far beyond it. Unlike the programs that Snowden exposed, this system does not operate in the shadows; it has public contracts, company names, and investment figures.

This perverse system does not end at the moment of detention. The photographs taken by Mobile Fortify are stored for 15 years; the contacts extracted from a detained person’s phone become new subjects of interest. And behind every stage, there is a company that profits. The Brennan Center calls it a “deportation industrial complex,” companies whose revenues depend directly on the development and growth of the automated system.

AI makes mistakes, like any automated system. But there is a fundamental difference between an AI trained to recommend movies and one trained on decades of data from U.S. security agencies to track down and deport people. When the former gets it wrong, you end up watching a movie you don’t like. When the latter makes a mistake, an innocent person can end up detained for weeks, like the Mexican migrant who made a living selling flowers. The problem is not just the error; it is that even when the system works as intended, a person can lose their freedom based on an algorithm’s decision, without ever being seen or heard, and without having any chance to defend themselves. The question that no agency has answered is who takes responsibility for this, and so far, there is no answer.

Palantir, Clearview, Thomson Reuters, LexisNexis, GEO Group: these are just a few of the companies that make up this ecosystem (there are dozens more that weren’t mentioned). What was described in this piece is only a fraction of a surveillance architecture that continues to expand, contract by contract, data point by data point. One question remains unanswered—one that should not be approached from a technological perspective, but rather from an ethical and political one: what kind of societies are being shaped by systems capable of depriving people of their freedom based on a score and a deportation date? Who are the first and primary victims of this automated system of expulsion?