July did not look like the winter raids. No tanks on local news loops. No secretary posing at a warehouse gate. The numbers still jumped.
ICE arrested 49,571 people last month, according to government files shared with the Deportation Data Project at UC Berkeley and UCLA. That is the highest monthly total since October 2022 and about 15 percent above June’s 43,021. Agents averaged roughly 1,500 arrests a day. Removals hit nearly 34,000 — about 1,100 people sent out of the country each day, the most since the second Trump term began.
Before inauguration, monthly arrests sat a little over 8,000, and most of those were jail transfers. December 2025 had already crossed 40,000. February dipped after the Minnesota surge and the political blowback that followed two fatal shootings. Then the line climbed again.
The raids got quieter. The machinery did not.
Kristi Noem’s tenure at DHS was built for cameras. Markwayne Mullin, who took over in March, has said the opposite out loud: keep the department out of the headlines and still hit the White House target of 2,000 arrests a day. A $70 billion funding package earlier this year gave the agencies more room to try.
Melissa Shepard, who runs legal services at ImmDef in Southern California, described the new pattern without much poetry. Fewer dozen-officer takedowns. More small teams at a traffic stop. People grabbed when they show up for a required ICE appointment. Others stopped outside their houses. Attorneys now tell clients with pending cases — even people cleared to travel — to think twice before flying, because airports have become another checkpoint.
Graeme Blair, co-director of the Deportation Data Project, put the same shift in operational language: arrests at airports, check-ins, and immigration courthouses, all at once. That mix, not a single spectacular raid, is what produced July.
Half of July’s arrests were non-custodial. In plain terms, ICE took people from the community rather than waiting for a jail handoff. Local partners mattered. AP’s read of the same ICE files found about 6,500 July arrests — roughly 13 percent of the month — came through state and local officers working under ICE. A year earlier those partnerships were under 5 percent. Texas and Florida together accounted for nearly 20,000 arrests. Texas, Florida, California, and Georgia produced more than half the national total. California is the outlier in that group; the other three lean hard on 287(g).
287(g) is the force multiplier nobody sees from the sidewalk
Section 287(g) lets ICE deputize local officers to do pieces of federal immigration work. Jail models question people already booked. Task-force models put that authority on patrol — traffic stops, calls for service, the kind of contact a bystander cannot easily tell apart from a regular citation.
The program was small for years. Trackers put it around 135 agreements in January 2025. By late August 2026 ICE’s own lists had climbed past 2,300 agreements across dozens of states. Florida’s sheriffs association has said every one of the state’s 67 counties signed on. Texas leads raw agreement counts. New York went the other way: a state law voiding 287(g) pacts took effect this week.
Mullin has called the partnerships “the greatest partnership we can have” and a way to work “in a more quiet way.” Volunteer rapid-response networks that trained last winter to film militarized raids are now rewriting their playbooks. A deputy with a local patch and a federal tasking sheet does not announce itself the same way a marked federal column does.
The “criminals first” line and what the files show
The administration still says the priority is people with criminal records. July’s arrest file does not match that speech.
The Guardian’s analysis of the same release found July was the first month of the second term in which a majority of arrestees — 51 percent — had no criminal conviction and no pending charge. Other write-ups of the AP analysis said more than half had neither a conviction nor a charge, and fewer than a quarter had a conviction. Detention snapshots from mid-July told a similar story: TRAC counted 65,765 people in ICE custody around July 11, with about 70 percent holding no criminal conviction.
None of that means ICE has stopped arresting people with records. It means the surge is not limited to that group. When daily volume becomes the metric, the pool widens to people who show up for an appointment, people with expired humanitarian status, and people who get pulled over in a 287(g) county.
Historical comparisons are messy because agencies mix removals, returns, and border turn-backs. Even so, the interior machine is running hot. ICE posted 356,389 removals for fiscal 2026 through mid-July. Obama-era peak years still sit near 410,000–438,000. At the current clip, this fiscal year is in that conversation, even if it remains well short of campaign talk about a million removals. Border crossings, meanwhile, have collapsed from the 2023–24 spikes, which is why so much of the new volume is coming from inside the country rather than the riverbank.
TPS is the next wave, and Springfield already felt it
Temporary Protected Status is a 1990 statute. Congress built it for people who cannot safely go home after war, disaster, or collapse. It is temporary on paper. In practice some designations lasted decades. El Salvador’s dates to earthquakes in 2001.
On June 25 the Supreme Court, 6–3, let the administration terminate TPS for Haiti and Syria. Justice Samuel Alito wrote that the statute bars courts from second-guessing those calls. Coverage put the Haiti population around 350,000 and Syria around 6,100. The administration had already ended or moved to end designations for a long list of other countries. El Salvador’s remaining protection — covering about 170,000 people — was set to lapse in September. Lebanon, Sudan, and Ukraine were among the last still on the calendar.
July’s arrest spike includes the first wave of Haitians after that ruling. Springfield, Ohio, where a large Haitian community settled in recent years, became a test site: people called into ICE offices, some fitted with ankle monitors. The first deportation flight to Haiti after the decision landed last week. Shepard’s line from Southern California travels well to the Midwest: she is “a little afraid for what it’s going to look like with the end of TPS.” She called it a precipice. That is lawyer-speak for a docket about to explode.
Advocates are telling TPS holders to file for every other form of relief that still exists. Many will not qualify. Work permits tied to the old designations have already started falling off employer I-9 files.
What the July record actually tells you
Three things sit in the same month.
First, volume. Nearly 50,000 arrests and nearly 34,000 removals is not a rounding error. DHS is closer to the 2,000-arrests-a-day goal than it was in February, even if it has not locked that pace in.
Second, method. The spectacle moved offstage. Check-ins, courthouses, airports, and local traffic units now do the work that winter raids advertised.
Third, legal status. A growing share of the people in those vans are not coming out of prison. They are coming out of protected-status cutoffs, missed paperwork windows, and ordinary contact with police in counties that signed a federal contract.
The data drop itself is a FOIA product, not a glossy DHS press kit. The Deportation Data Project released files through early August on August 24 and flagged the usual headaches: messy city fields, missing codes, removals tables that do not always match earlier versions. Treat the month-to-month comparisons as the best public window we have, not a perfect census.