Is ICE using medicaid data to arrest illegals?
Executive summary
A federal judge has allowed the Department of Health and Human Services (HHS)/CMS to resume sharing limited Medicaid enrollee data — basic biographical, contact and location information — with Immigration and Customs Enforcement (ICE), and reporting and court testimony indicate ICE is using tools that ingest that data to locate and arrest people believed to be unlawfully present [1] [2] [3]. Civil liberties groups, state plaintiffs and some news outlets warn the practice effectively turns health program records into enforcement leads, while courts have imposed explicit limits on the type and scope of information that can be shared [4] [5] [6].
1. The legal opening: judges and data-sharing orders
A series of court decisions in late 2025 and early 2026 resolved competing claims about whether HHS/CMS may share Medicaid information with DHS and ICE, with U.S. District Judge Vince Chhabria concluding that sharing “basic biographical, location and contact information” for people unlawfully present is authorized and may resume while litigation continues [2] [1] [3]. That ruling explicitly barred disclosure of sensitive medical records and limited sharing of data about U.S. citizens and lawful permanent residents, leaving a narrow but concrete legal pathway for ICE to obtain certain Medicaid-derived leads [3] [6].
2. What data can flow and how many people are affected
The Information Exchange Agreement and related CMS actions give DHS/ICE access to elements of the Transformed Medicaid Statistical Information System (T‑MSIS) and other biographical files that researchers say cover tens of millions of enrollees; reporting has cited figures in the range of tens of millions of Medicaid records that could be implicated by the exchange [5] [7] [8]. Public descriptions of the sharing emphasize that the data available for enforcement are primarily demographic, contact and address fields rather than clinical records, though advocates argue that even such fields are highly sensitive when used for enforcement [3] [5].
3. How ICE appears to be using the data — tools and vendors
Multiple outlets and advocacy groups report that ICE is using analytics platforms, including tools developed by Palantir such as ELITE (Enhanced Leads Identification & Targeting for Enforcement) and broader Palantir programs, to ingest Medicaid and other government data, map potential targets, produce “confidence scores” on addresses and guide enforcement operations [4] [9]. EFF and others characterize those tools as turning Medicaid enrollment information into arrest leads; media reporting and court testimony cited by 404 Media form the basis for claims about ELITE’s use of address and contact data to locate individuals [4] [9].
4. Limits, official positions and remaining defenses
Federal officials and some court language stress that the data-sharing is constrained — agencies say clinical records remain protected and Chhabria’s order narrowed the permissible set of information and barred sharing about citizens and lawful residents beyond narrow categories [3] [6]. The administration frames the step as lawful enforcement of immigration statutes, while states that sued argue it violates privacy laws and will chill health care access; that litigation remains active and subject to further judicial review [2] [8].
5. Consequences, controversies and what is not yet proven
Reporting documents immediate harms and fears — clinics and advocates report families skipping care, and civil‑liberties groups warn of a dragnet effect when benefit enrollment doubles as enforcement intelligence — but direct, public evidence linking specific arrests to particular Medicaid-sourced leads is still emerging in litigation and reporting [10] [11] [7]. Investigations and court filings indicate ICE has both the legal pathway and the technical means to use Medicaid-derived biographical data in deportation efforts, yet the full operational picture — including how often data from Medicaid specifically trigger arrests versus being one input among many — is not fully publicly documented as of the available reporting [4] [9] [1].