Two rates tell this story, and the gap between them is the finding.

Human Rights Watch and Physicians for Human Rights documented 52 deaths in ICE custody between January 20, 2025 and June 4, 2026. Thirty-nine came in the first twelve months - the highest annual total since ICE was created in 2003 [1].

The detained population also rose sharply: from roughly 40,000 to a peak above 71,000 in January 2026, easing to about 64,000 by April [1].

More people detained would mean more deaths even if nothing else changed. The raw count on its own proves little.

Here is why it is not just scale.

HRW reports the mortality rate rose 138 percent year over year, against a 77 percent rise in the detained population [1].

The rate grew at nearly twice the population's pace. Which means the chance of dying while detained went up - not merely the number of people exposed to it.

What happened to the reporting is the second half.

ICE has rescinded a 2021 policy that required a report when a former detainee died within 30 days of release. DHS described the rescission as 'common sense' [2].

Khaled Alrabe, an attorney with the National Immigration Project, named the incentive that creates: 'ICE can basically release people on their deathbed and then not have to worry about having to do any investigations' [2].

Whether or not anyone does that, the reporting line that would show it is gone.

One of the 52 has a name and a timeline.

Maksym Chernyak was 44. He died of a hemorrhagic stroke at the Krome North facility in Florida on February 20, 2025. His wife says he received no treatment for hypertension despite elevated readings, and that more than four and a half hours passed between the emergency call and his arrival at an emergency room [1].

One limit worth stating plainly: this newsroom could not independently verify ICE's own fiscal 2026 in-custody death count, because ice.gov returns a 403 to our checker. The figures here are the human rights organisations' documentation rather than the agency's.

That is a limit, and it is also the point. The count that is reliably available is the one produced from outside.