How ICE Uses AI to Automate Authoritarianism

Irna Landrum is a Minneapolis resident and senior campaigner on AI at Kairos Fellows.

The choppers overhead have not ceased for days. I sit on the phone monitoring my neighborhood's Signal chat because my neighbors have flagged heavy ICE and border patrol presence, abductions, possible door to door knocks, and lots of tear gas. It's on the block of my old house here in Minneapolis, where my 3-year-old nephew currently lives. I'm calling his mom, my sister-in-law, and giving her updates so their family doesn’t come home while ICE is still on a rampage. My nephew has already seen too much horror these past weeks.

Below the droning propeller wings, and amidst the chaos of community care, I am attempting to also do my day job as a campaigner focused on AI. I am witnessing ICE’s war against the people from the front lines, even as I study how this rogue militia is using cloud and AI technologies to automate authoritarianism. Because ‘authoritarianism’ as a term remains slippery, I adopt three critical parts of scholars Thomas Dekeyser, Casey Lynch, and Sophia Maalsen’s description of authoritarianism’s aims: it automates monitoring, expands repression, and centralizes power.

Before automating authoritarianism––that is, before automating monitoring, expanding repression, and centralizing power––ICE first consolidated massive troves of sensitive data. While original reporting suggested that it was using immigration-related data for its raids, we now know that ICE is hoarding information from an unprecedentedly diverse pool. These include both public and private databases alike, revealing information including our exact location, what we post online, and even private messages on our phones.

ICE bought this data from a vast array of players, from data brokers like Thomson Reuters and LexisNexis to utility companies to Big Tech corporations like Amazon. In other cases, ICE has simply appropriated data from government sources, like from the Department of Motor Vehicles in multiple states, or from federal programs like Medicare and Medicaid. The latest brick in the foundation for this massive, nonconsensual data set was laid by Elon Musk’s now-defunct “Department of Government Efficiency (DOGE),” which amassed data from the Social Security Administration, the Internal Revenue Service, and even voting records––all explicitly in order to track migrants, according to WIRED.

In addition to its sourcing, what ICE is doing with this data is also unprecedented. ICE is weaponizing cloud capabilities and AI not only to enable monitoring, but to automate this monitoring. It is developing a multi-source, continuous and real-time tracking tool against all of us. Again, DOGE laid the groundwork: it was reported last April that the rogue organization was hosting a “hackathon” for a “mega API” in the cloud that could pull information from real-time disparate sets of government data, making them effectively interoperable. WIRED reported then that such technology “could enable near-instantaneous access to tax information for use by DHS and immigration enforcement.”

It was also reported last year that Palantir is building a parallel tool for ICE, cruelly named “ImmigrationOS,” with a similar aim of funneling in different data sources, and developing frictionless, fast, and automated tools to build cases against those they want to terrorize. A parallel case exists in an ongoing genocide across the world: In Gaza, the Israeli Defense Forces “order from Amazon [Web Services]” when they need information about a Palestinian, according to +972 Magazine. In the US, ICE can likely do the same. After all, ICE and the IDF both use Amazon Web Services to power their respective surveillance machines.

ICE is also using AI to automate authoritarianism by expanding repression. Without technology, ICE is targeting dissidents like Renee Good or Alex Pretti outright: an explicit act of repression intended to chill dissent. With technology, ICE is automating this repression. An advocate from the Electronic Frontier Foundation (EFF) warned last year that the data consolidation tools described above could be used toward “repressive goals, to weaponize the information, use it against people [the state could] dislike, dissidents, surveil immigrants or other groups.”

That is no longer a hypothetical ‘could’. ICE and border patrol agents chose new targets with AI-powered surveillance, aggressing people beyond those they vilify as “illegals” to those “on work permits, asylum seekers, permanent residents, naturalized citizens, and even citizens by birth.” They are also targeting the poor, the working (and unemployed), and students and educators.

Ultimately, ICE is using cloud technology and AI to centralize power. In opposition to DOGE’s data distribution drive, Victoria Noble, a staff attorney at the EFF said, “There's a reason these systems are siloed.” Likewise, Cody Venzke, a senior policy counsel at the American Civil Liberties Union warned that this consolidation of data is barreling us toward a panopticon, or a “centralized dossier on everyone in this country.” An IRS worker added, "It's basically an open door controlled by Musk for all American's most sensitive information with none of the rules that normally secure that data." Though many such quotes reference Musk and DOGE, DOGE explicitly paved the way for ICE to automate its centralization of power.

In the face of a militarized occupation, when many of us are not leaving home for fear of being snatched in the streets, invoking authoritarianism is not far-fetched. And as my day job reminds me, AI, as it is currently marketed, developed, owned, powered, sold, and used, is being done so by corporations and the state toward authoritarian ends: to automate monitoring, expand repression, and centralize power. ICE is at the cutting edge of further advancing this disturbing trend.