ICE jails nearly everyone it arrests and deports most of them

Newly linked data show that under Trump, ICE booked about 90% of the almost 390,000 people it arrested in the interior into detention. That near-universal rate isn’t new, it had already climbed to nearly 90% by the end of the Biden years. But now ICE makes roughly triple the arrests, making detention both high-likelihood and high-volume. Removal is now the single most common outcome, with more than 261,000 people — about two-thirds of those arrested — ending in deportation or departure.

This analysis and writeup was generated by PromptQL, an AI-powered collaborative platform that makes sense of sprawling data sources, documents, and tribal knowledge. I wrote significant parts of it, edited it, reached out to sources, and reviewed the data analysis.

I’m consulting with the company and have found the platform is powerful and trustworthy enough to rely on for experiments like this and our 287(g) Watch state news summaries.

⁘ ⁘ ⁘

On July 13, 2026, an ICE Enforcement and Removal Operations officer in Biddeford, Maine, fatally shot a 26-year-old Colombian man who was authorized to work in the United States and was not the target of the arrest warrant — the second ICE deadly-force death in a week. It’s just the most recent example of an enforcement regime running full tilt.

On July 6, 2026, the Deportation Data Project released a new dataset linking each ICE arrest to what happened next. That matters because for the first time, a single anonymized, individual-level record follows a person from interior arrest through detention to removal. What it shows: almost everyone ICE arrests in the interior is jailed, and the most common outcome is deportation. Arrests roughly tripled after January 20, 2025, while the detention rate — already climbing through the Biden years — now sits around 90%.

DHS has issued more than 40 ‘worst of the worst’ press releases in the Trump era, touted a milestone of 10,000 gang members arrested, and celebrated 13 straight months of zero releases at the border. But the near-universal detention rate applies across the board, not just to the violent offenders their headlines feature.

In an email, a DHS spokesperson said, “The Deportation Data Project relies on information releases that have not been reviewed, audited or given context. DHS nor ICE have verified the accuracy, methodology or the analysis of the project and its results.” Despite saying they have not reviewed the data, the message goes on to claim: “The bottom line is that the Deportation Data Project is not accurate.”

Having not reviewed the information or methodology, they can’t know if it is accurate. The full response made further claims without providing evidence. It is available at the bottom of this post along with our methodology.

The Deportation Data Project does not systematically review third-party material pre-publication, but said they found no obvious inaccuracies in our analysis.

Almost everyone ICE arrests is now jailed

Of the almost 390,000 people ICE arrested since Trump’s inauguration, about 351,000 (90%) were booked into detention. That near-universal rate isn’t new — it had already climbed to nearly 90% by the end of the Biden years, up from a 61.3% period average that masks the rise. What changed under Trump is the volume: arrests roughly tripled. Combined, detention is now both high-likelihood and high-volume — ICE is arresting far more people and jailing almost all of them.

Open-ended custody is growing fast, too; almost 50,000 people arrested in this window remained in ICE custody, 14.2% of everyone detained. And people now cycle through more stints — a median of three detention stints per person, up from two.

What changed is not the speed or length of detention. Book-in still happens within a day, and the median completed stay is about 22 days in both eras. What changed is whether anyone is let go at all, and how many people are being arrested.

The most common outcome is removal

Earlier ICE arrest data published by the Deportation Data Project could not tell you whether a detention ended in deportation. This release links each arrest to its detention stay’s release reason and to the arrest record’s information about departed date, destination country, and final order. That makes the arrest-to-deportation path traceable per person.

These are outcomes as of the latest data, not a final tally. About one in eight Trump-era arrests — 12.8% — hadn’t resolved: the person was still in ICE custody when the records were pulled. And because the data lands in batches, the most recent months are incomplete. So the share below is a floor that will climb as open cases close — and some people counted as still in custody may already have been deported in ways the data hasn’t caught up to.

Even so, removal or departure is already the single largest Trump-era outcome, in a little more than two-thirds of all arrests — about 261,000 people. Mexico is the top destination, with almost 129,000 departures, followed by Guatemala at 42,000 and Honduras at 32,500; Venezuela, El Salvador, Ecuador, Nicaragua, and Colombia round out the leading countries.

Two-thirds of those arrested have no criminal conviction in the US

When DHS suggests that some of these people have criminal records in other countries, it may be right — but that is not what this data measures. ICE’s own classification counts only convictions in the United States, and by that measure two-thirds of the people it has arrested since January 2025 have none.

The largest single group ICE arrested since January 2025 is people the agency itself classifies as non-criminal “other immigration violators.” These people constituted 148,782 arrests, more than either the convicted or the charged. These are people with no criminal charge at all, meaning their only offense is a civil immigration matter, a pattern documented in depth in Marshall Project reporting on a Georgia traffic-stop arrest. Yet about 83% of them were detained, up from about 23% under Biden.

The outcomes for people with a criminal conviction and people with pending charges (but not conviction) was practically indistinguishable, with both hovering a little above 94%.

The composition has shifted month by month. In February 2025, non-criminal “other immigration violators” were 22.8% of arrests; by December 2025 they were 47.6%.

Local police partners extend immigration enforcement’s reach

The buildout runs partly through local law enforcement, which DHS has publicly courted. More than 1,720 state and local agencies now hold 287(g) agreements, led by Texas and Florida.

Those agreements surface directly in the arrest data, if modestly: about 20,000 arrests since Trump’s inauguration (roughly 5%) carry ICE’s 287(g) apprehension code, and 93% of them ended in detention, slightly above the overall 90% plateau.

However, that share is a floor, since arrests made inside local jails are frequently logged under other codes. Detention rates vary sharply by region, from 99.0% in the Newark, New Jersey area to 53.1% in Harlingen, Texas. DHS has repeatedly targeted the governors of holdout states, lodging detainers and demanding non-release in cases like an Illinois assault case and a Fairfax County, Virginia rape charge, and touting the removal of a man pardoned by Minnesota’s governor.

What these numbers do and don’t say

Everything here concerns ICE arrests and their consequences, not ICE detention as a whole. Many detention stays begin without an ICE arrest — most often at the border, following a CBP arrest — and are not included; a reader focused on detention itself should turn to the detention-stays datasets. A detention stay is attached only when book-in falls within five days before to ten days after the arrest, a 77% match rate, comfortably inside the range flagged as reliable. Months after February 2026 were incomplete and excluded.

And the criminality labels are ICE’s own administrative classifications, not court adjudications, a caveat that cuts against using them either to validate or to rebut the ‘worst of the worst’ framing.