I kept coming back to one number, and not because it was tidy. Forty percent. Not a rumor from a comment thread, not a campaign slogan polished until the edges disappeared. A sample review of asylum files tied to Afghan nationals resettled after the 2021 withdrawal found mistakes in roughly four out of every ten cases examined. That is the kind of figure that sits in your head while you make coffee. If a file is wrong two times out of five, what exactly did the system think it was measuring?
Maybe you have felt that same itch with any large administrative machine. A hospital chart missing an allergy. A bank form that never logged a second name. Annoying, usually fixable, rarely fatal. Immigration files are different. A missed alias or an unfinished security note is not a clerical bruise. It is a door left half open. I am not claiming every error hid a threat. I am saying the sample makes it harder to pretend the door was locked.
What Inspectors Actually Counted In The Sample
Oversight reviewers looked at a slice of adjudicated asylum applications filed by Afghan nationals brought in under the emergency resettlement effort that followed the withdrawal. The wider pool they drew from was 13,682 decided applications. From that pool they pulled 678 files. Inside that sample they logged a 40 percent error rate in how those asylum files were put together and closed.
The breakdown is more useful than the headline. Reviewers counted 303 missed aliases, 45 unresolved records, 23 missing security checks, and 115 other documentation problems. Most of the trouble, they said, came from aliases that never made it into the file or from paperwork that simply was not finished. Two cases sat in a narrower, sharper category: potential inadmissibility tied to terrorism-related grounds.
Pause on the aliases for a second. A person can carry more than one name for ordinary reasons. Marriage. Transliteration. A nickname that became official somewhere along a border. A file that does not list those names is a file that cannot be matched cleanly against older records. I have watched smaller organizations trip over this. Multiply it by a rushed national program and the miss stops looking like a typo.
The Two Cases That Should Have Slowed Everyone Down
The terrorism-related errors were not described as confirmed plots. They were process failures around a specific legal screen known as Terrorism-Related Inadmissibility Grounds. In one file, an asylum officer did not place the required worksheet in the case record. In the other, officials never marked whether they accepted or rejected the officer’s analysis of an applicant who was subject to that screen.
That second miss is the one that bothers me more. A worksheet left out is sloppy. A decision left unmarked is a conversation that never finished. Someone looked at a possible bar to admission and the file does not say what the institution concluded. Years later, a new officer opening that folder would be guessing. Guessing is a poor substitute for a record.
Most errors resulted from missed aliases or incomplete documentation, including two cases related to potential inadmissibility on terrorism-related grounds.
Oversight review of sampled asylum files
Read that sentence twice. The reviewers were not announcing a wave of missed terrorists. They were announcing that the paper trail around two sensitive analyses was broken. In a system that lives or dies on paper trails, a broken trail is the finding.
How The Sample Sits Against The Full Caseload
A sample is not the whole story, and anyone who treats 678 files as if they were all 13,682 is overselling the math. Still, 678 is not a handful. It is large enough that a 40 percent error rate is difficult to wave away as a bad week in one office. If the sample was drawn to represent adjudicated files from that population, the burden shifts. The agency has to explain why the rest of the pile would look cleaner.
Here is a plain way to see the categories the reviewers named.
| Issue flagged | Count in sample | What it usually means |
| Missed aliases | 303 | Other names not captured for matching |
| Other documentation issues | 115 | Incomplete or inconsistent file contents |
| Unresolved records | 45 | Hits or notes left without a clear close |
| Missing security checks | 23 | A required screen not shown in the file |
| Terrorism-grounds process errors | 2 | Worksheet missing, or approval mark absent |
The table is a snapshot, not a criminal docket. Categories can overlap in a single file, which is why the counts do not need to add neatly to 40 percent of 678. The point is the shape of the mistakes. Aliases dominate. Security-check gaps are fewer, and more serious when they appear.
Why Aliases Are Not A Minor Clerical Habit
Think of an alias the way a librarian thinks of a cross-reference. Without it, the book is in the building and still impossible to find. Biographical screening depends on names, dates, and places lining up across databases. Drop a name and you drop a possible match to an older warrant, a prior visa refusal, or a watchlist entry filed under a spelling no one in the new office would guess.
Transliteration from Dari or Pashto into Latin letters is famously unstable. One person’s name can land on paper three different ways before anyone intends deception. A careful file anticipates that. A rushed file records the version on the form in front of the officer and moves on. The review suggests a lot of files took the second path.
The Program The Files Came From
The applications belonged to people resettled through the effort commonly known as Operation Allies Welcome, stood up after the withdrawal from Afghanistan. The stated aim was to bring in roughly 97,000 Afghan nationals. Airports filled with families. Processing centers ran around the clock. The political promise, repeated often at the time, was that people who had worked with the United States, and who had passed serious screening, would be protected.
That promise had a moral core. Interpreters, drivers, and contractors who had stood next to American personnel were in real danger once the government collapsed. Leaving them behind would have been an ugly ending to a long war. The trouble is that a moral core does not adjudicate a file. Speed and sympathy can share a room with rigor. They do not replace it.
Later research summaries have argued that the population which actually arrived was broader than the ally narrative. One December 2025 review put the number of Afghan nationals who entered under the prior administration’s post-withdrawal policies above 200,000, and said most had no tie to the government or its contractors. They were not, in that account, uniformly former partners, nor uniformly people already found to be persecuted refugees. If that reading is even partly right, the vetting story was sold in a narrower frame than the caseload that showed up.
A Deadline That Staff Said Would Bend The Work
Reviewers did not open this evaluation on a whim. They started after complaints that a settlement would warp how asylum cases were decided. In September 2023 the department agreed with certain plaintiffs to adjudicate 90 percent of asylum applications from this resettlement population within 150 days. Ordinary applications from outside that group can run to 180 days. Fifteen days sounds small until you are the person holding the file.
Employees told reviewers the shorter clock could hurt a proper read. That complaint is the least surprising sentence in the whole report. Anyone who has worked a queue knows what a forced deadline does. Corners round off. Notes get thinner. The item that takes an extra phone call gets a shrug and a stamp. I have seen it in newsrooms and in claims offices. Immigration adjudication is not magically immune.
- Standard outside cases could take up to 180 days.
- The settlement pushed 90 percent of this population’s cases toward 150 days.
- Staff warned that the squeeze could weaken review quality.
- Reviewers later found a 40 percent error rate in the sample they pulled.
Cause and effect are not proven by placing those bullets next to each other. A deadline can coexist with sloppy files that would have been sloppy anyway. Still, when workers predict a quality drop and a later sample shows a quality drop, pretending the clock was irrelevant is a stretch. Perhaps the most interesting aspect is how little institutional memory seems to have been built around that warning before the files were closed.
What The Agency Said In Its Defense
The citizenship and immigration service answered a redacted draft. Its director wrote that the draft failed to weigh missing aliases and imperfect security checks against broader screening already in place. Those safeguards, the letter argued, include biometric and biographical checks against multiple law enforcement and watchlist databases. The agency also said it keeps tightening vetting for people who apply for asylum.
That reply is not empty. Fingerprints and database queries can catch what a missing alias misses, if the biometrics were actually taken and if the databases contain the right record. A worksheet left out of a paper file does not automatically mean the person was never run through a system. I grant that. File hygiene and security screening are related, not identical.
The reply still leaves a hole. If the safeguards were as robust as claimed, why did 23 files in a 678-file sample show missing security checks at all? Why did two terrorism-grounds analyses fail the basic test of being fully recorded? Robust systems are supposed to make those misses rare and visible, not common enough to headline an oversight report. A defense that says “the net is bigger than the hole” has to show the net was actually cast on the files with holes.
Instances of missing aliases or improper security checks were taking place amid broader, established, and robust screening, vetting, and security checks safeguards already in place that minimize the impact of these errors.
Agency response to the draft oversight review
“Minimize” is a careful verb. It does not mean erase. It means the agency believes the leftover risk is small. Belief is not the same as a re-review of the 303 alias misses. Until those files are pulled back and matched again, the minimization is an argument, not a result.
The Public Safety Cases Officials Later Highlighted
After the review circulated, department leadership said the resettlement effort had allowed people later accused of serious crimes to enter and remain. The list they offered included sexual battery, kidnapping, drug possession, hit-and-runs, and lewd or lascivious acts involving a minor. Those are accusations and case descriptions from officials, not a court summary I can independently retry here. They are still the sort of outcomes a vetting system is built to reduce.
A few episodes were dated with uncomfortable precision. In November 2025, the department helped federal investigators arrest an Afghan national charged with threatening to kill Americans. He had come in through the same operation. In December 2025, a member of the Afghan affiliate of the Islamic State group, also admitted under the operation, was arrested. In March of this year, an Afghan national allowed in under the program, previously convicted of indecent exposure to a minor, was arrested.
Three cases do not indict tens of thousands of people. Anyone who slides from “these arrests happened” to “the population is dangerous” is doing bad math and worse politics. The honest use of the cases is narrower. They show that at least some individuals who should have been hard stops were not stopped at the gate. A system that misses a prior conviction involving a minor, or that admits someone later tied to an extremist affiliate, has a calibration problem. Scale does not wash that out.
How Leadership Framed The Failure
The current homeland security secretary called the asylum system something that should be as strict as possible, and said a single dangerous criminal or terrorist who slips through endangers a large number of lives. He described the resettlement program as dangerously mismanaged, and said it set aside thorough, established asylum procedures. That is political language. It is also a fair description of a settlement that shortened the clock and a sample that came back full of unfinished notes.
Earlier, in June 2022, the same department under the prior secretary had said Afghans who supported and worked with the United States, and who had undergone rigorous vetting, could qualify for protection and other benefits. The secretary at the time said the approach would allow eligible people who posed no national security or public safety risk to receive asylum, refugee status, or another lawful status. The contrast between “rigorous” in 2022 and “40 percent error rate” in the later sample is the whole argument, compressed.
Both statements can be true in pieces. Some applicants were carefully screened. Some files were not. A program can contain real allies and real process failures at the same time. The political habit is to pick one half and throw the other half away. The files do not cooperate with that habit.
What “Rigorous Vetting” Was Supposed To Include
People throw the phrase around until it means whatever the speaker needs. In practice, screening of this kind usually stacks several layers. Identity documents. Biometrics. Name checks against law enforcement holdings. Watchlist queries. An interview. A legal analysis if a terrorism-related ground might apply. Then a recorded decision. Skip a layer, or skip the record of a layer, and the stack is shorter than the brochure.
In my experience reading oversight write-ups, the layer that fails first is rarely the dramatic one. It is the alias line. It is the checkbox. It is the worksheet that lived on someone’s desk and never reached the folder. Drama arrives later, when a name that should have matched finally matches, or when an arrest makes the missing checkbox famous.
A workable screen, stripped of slogans: Identity that can be checked Names, including the awkward ones Biometrics actually collected Databases actually queried Hits actually resolved Sensitive grounds actually written down A decision a stranger can audit
That list is not exotic. It is the minimum a skeptical auditor would expect. The sample failed pieces of it often enough to hit 40 percent. That is the finding worth sitting with, before anyone reaches for a larger theory about borders or wars.
Speed As A Kind Of Policy
There is a metaphor I cannot shake. A fire exit with a broken latch is still a door. People will use it because the building is loud and the clock is running. The resettlement effort was a fire exit. Kabul fell. Flights left. Processing centers became the only available room. In that setting, “faster” felt like mercy. Mercy without a latch is how you get a report like this one.
None of that erases the debt owed to people who actually worked alongside American forces. A driver who spent years outside the wire should not be punished because a later applicant’s alias was missed. The policy error was blending populations, compressing timelines, and then speaking as if every file had received the same careful pass. Blending is easier than sorting. Sorting is the job.
Would a 180-day clock have caught the 303 aliases? Maybe not all of them. A longer clock does not create competence. It does create room for a second look, and the staff who complained seemed to want exactly that room. Ignoring them was a choice. Choices have samples attached to them later.
What A Re-Review Would Need To Prove
If the agency wants the public to accept the “safeguards minimized the impact” line, the next step is boring and specific. Pull the error files. Re-run the alias matches. Show the biometric queries that existed even when the paper was thin. Mark the two terrorism-grounds cases with an actual approve-or-disapprove note, and say what that note rests on. Publish the count of files where a miss changed the outcome, not just the count of misses.
Until that happens, readers are left with two stacks of paper that do not quite touch. One stack says the sample was full of holes. The other stack says the holes did not matter because other nets were tight. Both stacks can be cited. Only a re-review can tell you which one was describing reality.
- Identify every sampled file with a missed alias or unresolved record.
- Re-query those names and biometrics against current holdings.
- Close the two terrorism-grounds files with a written decision.
- Report how many outcomes change, including how many do not.
- Apply the same pass to files outside the sample if the hit rate is high.
That sequence is not radical. It is what a cautious records office does after an auditor finds a pattern. The absence of a public sequence like it is, to me, more telling than another press statement.
Numbers That Get Misused On Both Sides
Forty percent is easy to weaponize. One camp will say it proves the program was a sieve. Another will say file errors are not people, and that equating a missing alias with a terrorist is smear. Both camps skip the middle, which is where the report actually lives. The middle says documentation failed at a high rate, security-check gaps were real but smaller, and two sensitive analyses were not properly recorded. That middle is less shareable. It is also the part a serious reader should refuse to leave.
The 97,000 figure and the later 200,000 figure will get the same treatment. One is the early program aim. The other is a broader count of entries under post-withdrawal policies. They answer different questions. Mixing them into a single scare number, or a single reassurance number, is how debate gets dumber. I would rather keep them in separate columns.
Arrest anecdotes have a similar gravity problem. They are concrete, so they feel like proof of a trend. A trend needs rates. How many entrants, over what period, produced how many charged offenses, compared with what baseline? Officials have not handed the public that table. Without it, the anecdotes prove possibility. They do not prove proportion. Possibility is already enough to justify fixing the files. It is not enough to justify collective blame.
The Human Cost Of A Thin File, Either Way
A bad file can fail in two directions, and both are ugly. It can admit someone who should have been stopped. It can also trap someone who should have been cleared, because an unresolved record sits there like a stone and no one picks it up. The 45 unresolved records in the sample are not automatically villains. They are unfinished questions. Unfinished questions delay lives. Families wait. Work authorization stalls. The same sloppiness that frightens security hawks also punishes applicants who did everything asked of them.
That double edge is why “just go faster” and “just say no” are both lazy. Faster produced this sample. A blanket no would have abandoned people who really had been targeted for working with foreign troops. The adult version of the policy is slower where the record is dirty, faster where the record is clean, and honest about which is which. Honesty requires the worksheet to be in the folder.
What Ordinary Readers Can Reasonably Conclude
You do not need a law degree to hold a few conclusions at once. The sample error rate is high. Alias capture was the weak joint. A settlement clock made staff nervous, and the nervousness looks justified in hindsight. Agency leaders say other checks limited the damage. A handful of later arrests show that damage was not purely theoretical. Most people in the program are not those arrests. The ally story was cleaner than the arrival story some researchers describe.
If that feels unresolved, good. It should. Oversight is supposed to leave a residue of unfinished work, not a victory lap. The residue here is a pile of files that still need a second pass, and a public argument that keeps substituting volume for verification.
I keep returning to the unmarked analysis. Somewhere an officer wrote something about a possible terrorism-related bar. Somewhere a supervisor was supposed to say yes or no. The file does not show the answer. That is a small silence with a large echo. Systems fail in silences like that, not only in the speeches that come afterward.
Questions The Next Review Ought To Answer
A useful follow-up would be dull on purpose. Did error rates differ by office, by month, or by how close a case sat to the 150-day line? Were missed aliases more common when names had multiple transliterations? Did the 23 missing security checks correspond to people who also lacked biometrics, or only to people whose biometrics existed outside the asylum folder? How many of the sampled applicants have since had a law-enforcement encounter, and how does that rate compare with other asylum cohorts?
Those questions are answerable. They are not answered by calling the program a success or a catastrophe. They are answered by opening folders. Until the folders are opened again in public view, the 40 percent figure will keep doing all the talking, which is more work than one number should have to do.
Sample logic, in one line: 678 files reviewed, error patterns in about 40 percent, aliases first, security gaps fewer, two terrorism-grounds records incomplete.Carry that line, not the slogan that gets built on top of it. Slogans age badly. File counts do not.
A Closing Look At The Gap Between Promise And Folder
The promise in 2022 was protection for eligible people who posed no national security or public safety risk, after rigorous vetting. The folder, in the sample reviewed later, often did not look rigorous. Between those two sentences is the actual story. Not a myth about an invasion. Not a myth about a flawless rescue. A story about a large administrative promise that outran its paperwork.
Paperwork sounds small next to a war’s end. It is not small to the person who relies on it, whether that person is an applicant waiting for a fair read or a resident who assumes the gate was watched. Both are owed the same unglamorous thing: a file that can survive a stranger’s audit. On the evidence of this sample, too many files would not.
Fixing that does not require a new speech. It requires the aliases, the unresolved records, the missing checks, and the two unmarked analyses to be worked until a second reviewer can follow them without squinting. Anything short of that is another announcement. Announcements are what created the confidence this report just punctured.
