The Trump administration is sharply expanding its use of artificial intelligence in immigration enforcement, using technology not just to track migrants but also to help determine who gets targeted for deportation.
To power immigration enforcement — a top policy priority for President Donald Trump — the administration is deploying artificial intelligence algorithms to sift through a vast array of records. Officials say the tools can flag potential violations, prioritize leads and direct officers on next steps, accelerating processes that once relied on slower, manual reviews.
Acting Immigration and Customs Enforcement Director Todd Lyons has described a vision of squads of trucks sweeping up immigrants with the efficiency of Amazon delivery routes: “Like Prime, but with human beings,” he said in April during the Border Security Expo in Phoenix.
Driving that vision is ImmigrationOS, a new platform that consolidates these tools into a single interface. The system, which DHS will start using Thursday, includes workflows that allow agents to approve raids, book arrests, generate legal documents and route individuals to deportation flights or detention — all in one place.
“It doesn’t just collect data — it structures what agents do with it,” a senior DHS official said.
While some of the technology has been previously used piecemeal for immigration efforts, the scale of this project is unprecedented. The system also draws from traditionally non-immigration data sources, including Suspicious Activity Reports and financial transactions flagged under the Bank Secrecy Act. Those tools — more commonly deployed in counterterrorism or anti-money laundering cases — are now being repurposed to identify potential immigration enforcement targets, from people suspected of identity fraud to those working without authorization.
Some experts warn the growing reliance on opaque algorithms raises serious concerns. Bias, overreach and reduced human oversight are all possible, particularly as the Department of Homeland Security ramps up deportations and executive orders pushing AI adoption while dismantling so-called woke or regulated systems.
“What we’re seeing in these upcoming contracts is a shift,” said Steven Hubbard, a data scientist at the American Immigration Council, a pro-immigration nonprofit. “In the past, AI tools were primarily used to support efficiency — flagging a case for review or assisting with facial recognition. But the new systems described appear to be more prescriptive, guiding agents on what actions to take rather than simply supporting their decision-making.”
Under the Biden administration, DHS began publishing AI “inventories” to disclose how agencies use artificial intelligence. These reports reveal that AI tools are already embedded in identity verification, fraud detection, check-ins for alternatives to detention, and even chatbots for internal use.
Now that effort is ramping up, with the help of Palantir, a major government contractor also used by the Department of Defense.
Earlier this year, ICE awarded the company nearly $30 million to develop ImmigrationOS. The system is being designed not just to collect information but also to shape enforcement decisions — a move critics say outsources judgment to software and raises concerns about oversight and accountability.
CNN has reached out to DHS and Palantir for comment.
Palantir has said that it believes AI should be used to “enhance human execution and decision-making rather than replacing it” and that “it’s essential to preserve fundamental principles of privacy and civil liberties while using data.”
A former DHS official who worked in the current Trump administration told CNN that ImmigrationOS marks a fundamental shift in enforcement infrastructure. “They’re combining data sets from Treasury and other agencies into an observability platform,” the official said.
The former official added that the new platform will also use Internal Revenue Service and census data. It includes built-in workflows that allow agents to approve raids, book arrests, generate legal documents, and route individuals to deportation flights or detention — all from a single interface.
“You can go in and, you know, raid and arrest, get your approval to do that in the system,” the official said. “Then it’ll route them through to detention or removal. That entire lifecycle is built into the software.”
The former official also raised alarms about ICE’s growing dependence on Palantir. “One of the things that we’ve been concerned about is relying so heavily on a primary vendor … it means that you’re locked into their ecosystem,” they said. “Substantial portions of the IT infrastructure now require Palantir’s tools to function.”
Beyond vendor lock-in, the system also marks a decentralization of authority. Under new Trump-era directives, ImmigrationOS will be accessible directly to officers on the ground — not just at headquarters. “The idea is to push AI down to the operators,” the official said. “Let the officers on the ground use it how they see fit, instead of managing it centrally.”
Hubbard cautions that while the focus is often on facial recognition or surveillance, the real impact may lie deeper — in the quiet, automated decisions that affect who gets targeted, detained or deported. “If AI becomes the basis for enforcement action,” he said, “oversight isn’t optional — it’s urgent.”
Former acting ICE Director John Sandweg, who served during the Obama administration, says the debate isn’t about whether AI can be useful, but how it’s deployed. “These tools can be used for good or bad,” Sandweg said. “This administration has made clear it wants to ramp up arrests — they’ve said it themselves.”
Sandweg noted that ICE’s embrace of technology is not new. “AI tools have been around for more than a decade. They can play a critical role in identifying threats and terrorists. ICE has always embraced technology — in the past, it was to focus on terrorism threats.”
But he questioned the need for AI in certain day-to-day enforcement operations. “You don’t need AI to go to Home Depot and arrest people,” he said.
Additionally, a former top DHS official, speaking at an immigration compliance panel under Chatham House rules, said ICE “now has a new AI tool to do the work of I-9 auditors more quickly and effectively.” Such auditors are responsible for reviewing employment eligibility forms (known as I-9s) to verify that workers are legally authorized to work in the US. They check documents such as passports, green cards and visas, and flag discrepancies or potential violations.
But critics say that same efficiency, if left unchecked, could sideline human judgment in favor of algorithmic speed — with real-world consequences for real people.
“You can’t take it [AI] as gospel,” said a different former DHS official. “AI is good, but not as good as a human being.”