How many people died in ICE custody each year since 2000 according to ICE’s official reports?
Executive summary
ICE publishes individual detainee death reports and a death-reporting policy, but the sources provided do not contain a single, official year-by-year tally from 2000 through 2025; available reporting therefore permits confirmation of some headline years (notably 2004 and 2025) and trends (COVID-era peaks and more recent rises), while the exact annual counts for every year since 2000 must be obtained directly from ICE’s Detainee Death Reporting archive (ICE) [1] or reconstructed from the agency’s individual death reports [2].
1. What the agency officially publishes and what that means for a year-by-year count
ICE’s public-facing process requires local field directors to notify ICE leadership and to post detainee death reports; the agency states it posts report summaries and notifies Congress and stakeholders when a death occurs, and a 2021 directive formalized notification, review and reporting requirements (ICE) [1], (ICE release examples) [2]; however, none of the supplied sources contains a consolidated table listing the number of ICE-custody deaths for every year since 2000, meaning an authoritative per-year total must be built from ICE’s archived Detainee Death Reporting entries [1].
2. Confirmed headline years from the supplied reporting
Multiple news outlets and watchdogs agree that 2025 was the deadliest recent year: Reuters reported “at least 30 people died in ICE custody in 2025” and noted the figure was the highest in two decades (Reuters) [3], while The Guardian and other coverage placed the 2025 toll at 32 and said that number matched the previous record set in 2004 (The Guardian) [4], (New Republic) [5], indicating strong corroboration among independent outlets for a 2004/2025 peak in the low‑30s.
3. Broader multi‑year signals and the COVID‑era peak
Academic and public‑health analyses show a clear COVID-era spike in deaths: peer‑reviewed reporting and reviews of FY2018–2023 note an elevated death rate during FY2020 driven by COVID‑19 and related causes, with subsequent declines through FY2021–2023 (NIH/PubMed) [6]; advocacy organizations and investigative journalists have documented clusters of deaths and criticized medical care during those years (ACLU analysis of 2017–2021 deaths) [7].
4. Why publicly available counts can differ and why some deaths might be omitted from quick tallies
Independent trackers note several reasons counts can diverge: ICE historically posts narrative death reports (sometimes under euphemistic headlines) and, per secondary reporting, has been accused of undercounting deaths that occur shortly after release or that are not reflected in agency press releases—an issue raised by advocates and documented reporting (Wikipedia/advocacy notes) [8], (ACLU/DWN report) [9]; the 2018 DHS appropriations language also imposed timelines for public reporting, which improved transparency but did not eliminate disputes about classification (Wikipedia) [8].
5. What can be stated with confidence from the supplied sources
From the material provided, it can be asserted with sourced confidence that 2025 saw the highest number of ICE‑reported in‑custody deaths in roughly two decades—reported as at least 30 by Reuters and as 32 by The Guardian and other outlets—and that this matched the previous high recorded in 2004 [3], [4], [5]; beyond those headline years, academic reviews confirm a COVID‑era peak around FY2020 with reductions afterward, and advocacy reports document many preventable deaths in the 2017–2021 period [6], [7].
6. The responsible next step to obtain an exact, year‑by‑year list from 2000–present
For an authoritative, year‑by‑year accounting “according to ICE’s official reports,” the only reliable path is to extract and aggregate ICE’s individual Detainee Death Reporting entries hosted by the agency—ICE’s Detainee Death Reporting page is the primary source named in ICE statements and press releases [1], [2]—or to request an official dataset from ICE or DHS under FOIA if the archived public entries are incomplete; the supplied reporting does not itself enumerate every annual total since 2000, so this aggregation step is essential to answer the question exhaustively.