How many total arrests did ICE make nationwide in 2025 and how are they categorized?
Executive summary
ICE arrest totals for calendar 2025 vary by dataset and definition, but major compilations assembled from ICE releases and FOIA disclosures place the agency’s arrests in the low hundreds of thousands — The Guardian reports “more than 328,000” arrests for 2025 (to mid/late December) based on ICE data [1], while FOIA-based analyses and news outlets describe an operational tempo of roughly 1,000–1,100 ICE arrests per day through much of the year [2] [3]. Those arrests are reported and categorized unevenly across multiple dimensions: arresting agency and program, arrest location (including a large share from local jails), and criminal-history buckets that ICE and DHS use to distinguish “criminal” from noncriminal arrests [4] [2] [5].
1. What the headline numbers show: total arrests reported in 2025
Public compilations of ICE’s detention-management and FOIA-provided arrest releases put the count of people arrested by the administration in 2025 at roughly the low-to-mid 300,000s; The Guardian’s ongoing scrape of ICE releases and other government feeds calculates “more than 328,000” arrests through mid‑December 2025 [1], and FOIA data analyzed by groups such as the Deportation Data Project and reporters show an average arrest pace of about 1,000–1,100 per day during large parts of the year [2] [3]. Those daily averages align with media reconstructions but do not map perfectly onto a single definitive ICE “annual” total because ICE’s public tables and FOIA releases cover different date ranges and counting rules [6] [7].
2. How ICE categorizes arrests: agencies, locations and “custodial” coding
ICE and its analysts break arrests down by the arresting subagency (Enforcement and Removal Operations/ERO, Homeland Security Investigations/HSI, and inputs from CBP and local partners), by the location of arrest (streets/at‑large, workplaces, homes, courts, and local jails or lock‑ups) and by program labels such as the Criminal Alien Program (CAP) or, more recently, a generic “Custodial Arrests” category [4] [2]. Reporting practices shifted mid‑2025 when ICE stopped separately flagging many CAP arrests and instead lumped numerous jail‑origin cases under “Custodial Arrests,” a change that obscures earlier distinctions between arrests initiated in state/federal prisons, local jails, or via CAP partnerships [2].
3. Criminality buckets and competing narratives
ICE’s public statistics and DHS statements present arrests categorized by criminal history — convicted, pending charges, or no criminal conviction — and DHS has emphasized that a majority of arrests involved people with criminal convictions or charges, with officials asserting roughly 70% criminality in ICE arrests [4] [5] [8]. Independent analyses and watchdog reporting complicate that frame: some outlets and researchers show that large portions of detention growth and many arrests have been of people without convictions or only for nonviolent offenses, while the DHS “worst of the worst” database represents a small fraction of total arrests (a contested 4% figure cited by critics) [9] [10] [11].
4. What to watch in the numbers — limitations and changes that matter
The headline totals are shaped by definitional choices: ICE’s public “Initial Book‑Ins by Arresting Agency and Month” counts arrests that lead to book‑ins or detention and can undercount at‑large contacts that don’t enter ICE custody; FOIA datasets released via the Deportation Data Project extend transparency but currently cover through mid‑October in their última release, leaving year‑end aggregation to third‑party scrapes [7] [6]. Crucially, reporting changes during 2025 — especially ICE’s relabeling of many CAP/jail arrests as “Custodial Arrests” and the agency’s mix of arresting units — mean that simple headline totals (e.g., ~328,000) must be read alongside metadata about who counted what, where those arrests occurred (nearly half concentrated in local jails per one FOIA analysis), and which subset are classified as criminal versus noncriminal [2] [6].