Immigration to the United States and customs law enforcement exploits by Palantir generative artificial intelligence tools to sort and summarize immigration application guidance from its public submission form, according to an inventory released Wednesday of all Department of Homeland Security use cases for AI in 2025.
The AI Enhanced ICE Tip Processing service is intended to help ICE investigators “more quickly identify and implement tips” for urgent cases, as well as translate submissions that are not written in English, according to the inventory. It also provides a “BLUF”, defined as a “high-level summary of the trick”, produced using at least one large language model. BLUF, or “bottom line up front”, is a military term used also used internally by certain Palantir employees.
DHS says the software is “actively authorized” to support ICE operations, adding that the tool helps reduce “the tedious manual effort required to review and categorize incoming tips.” The date AI-enhanced tip processing “became operational” is listed in the inventory as May 2, 2025.
The DHS inventory does not provide much detail on the broad language models Palantir uses to generate BLUFs; however, it notes that ICE uses “large, commercially available language models” that have been “trained on public domain data by their vendors.”
“There has been no additional training using agency data beyond what is available in the models’ core capability set,” the inventory also notes. “During operation, AI models interact with submitted tips.”
The “DHS 2025 AI Use Case Inventory” published Wednesday on the DHS website, has been published annually since 2022. The 2024 version of the inventory does not mention the use of AI to process tip line submissions.
Palantir has been a major ICE contractor since 2011 and provides a comprehensive set of analytical tools to the agency. However, until now, almost nothing was known about Palantir’s advice on processing work for ICE.
This work was once mentioned in the description of a $1.96 million Palantir. payment that ICE made in September 2025. The payment was to modify the Investigative Case Management (ICM) system—a version of Gotham, Palantir’s out-of-the-box law enforcement product that stores information on current or past ICE investigations, to include the Tipline and Investigative Leads Suite.
The description does not include any other details about Palantir’s work on this “Tipline” integration.
However, the “AI Enhanced ICE Tip Processing” tool may be an update to “FALCON Tipline”, which replaced ICE’s old tip processing system around 2012.
Palantir, ICE and DHS did not immediately respond to requests for comment.
According to a DHS document last updated in 2021the FALCON Tipline processes information submitted by the public or law enforcement agencies regarding “suspected illegal activity” or “suspicious activity” to ICE’s Homeland Security Investigations (HSI) Tipline Unit. ICE appears to have only one reporting line, but submissions can be made online or by phone.
An entry in a federal register in December 2025 notes that when HSI receives a tip, investigators in its Tipline unit run “queries” against various “DHS, law enforcement, and immigration databases.” After analyzing these results, HSI agents write “investigative reports” and then forward the information to the appropriate DHS offices. It is unclear to what extent this workflow may be facilitated by recently enhanced AI processing.
Data from FALCON Tipline, Palantir ICM and several other databases are ingested and made searchable by the FALCON Search and Analysis System, a separate but similarly named tool also developed by Palantir.
After the federal agents shot and killed Minneapolis nurse Alex Pretti On Saturday, Palantir workers pressed executives for answers about the company’s work with ICE. In Slack messages, reviewed by WIRED this weekworkers asked if Palantir could “put pressure on ICE.” One worker wrote: “Our involvement in ice cream has been overlooked too much internally under Trump2. We need to understand our involvement here.”
In response to this pressure, management updated Palantir’s internal wiki detailing its ongoing work with ICE. In a Jan. 24 post, Akash Jain, whose LinkedIn profile lists him as chief technology officer and president of Palantir USG, defended the company’s work with ICE, writing that Palantir’s services improve “ICE’s operational efficiencies.”
“There have been growing and increasingly visible field operations focused on domestic immigration enforcement that continue to draw attention to Palantir’s involvement with ICE,” the wiki says. “We believe our work could have a real, positive impact on ICE’s enforcement operations by providing officers and agents with the data needed to make more accurate and informed decisions. We are committed to providing our partners with the best software for their work, while recognizing the reputational risk we face when supporting immigration enforcement operations.”
The updated wiki describes Palantir’s work with ICE as focusing on three main areas: “Prioritization and targeting of enforcement operations,” “Self-deportation tracking,” and “Immigration lifestyle operations focused on logistical planning and execution.” But it doesn’t mention any use of AI to help immigration enforcement agents sort through potential information.
The inventory released by ICE on Wednesday also refers to another tool developed by Palantir, called Enhanced Leads Identification & Targeting for Enforcement (ELITE), which was first reported by 404 Media earlier this month. ELITE creates maps describing potential deportation targets and presents informational files on each person. The tool pulls data from the Department of Health and Human Services (HHS) to identify addresses of potential targets. The tool became operational in June, according to the inventory, and 404 Media reports it has been used in Oregon.
“While ELITE provides actionable data to Enforcement and Removal Operations (ERO) officers, its results are limited to standardized address data and do not serve as the primary basis for decisions or actions that have legal, material, binding, or significant effects on individuals,” the inventory states. “ICE data was not used during the design, development, or training phases of the AI models. During operation, the AI models interact with ICE production data from multiple sources, including data from ICE’s Integrated Application Database (EID).”
ICE and the White House have repeatedly turned to the agency’s web form for tips over the past year, calling on the public, not just law enforcement, to submit possible leads. “Help ICE agents make your community safer by reporting suspicious activity,” said an ICE post on X from February.
