1. A working telephone, on paper
The telephone was supposed to connect a detained person to someone outside the detention facility. Among the numbers available was the Department of Homeland Security inspector general's hotline, a route for reporting problems to the office charged with investigating the department itself.
When federal auditors tried the toll-free numbers during a detention inspection, they could not get through. The telephone company's restrictions prevented the calls. The auditors told facility employees, who worked on changing the dialing instructions. According to the auditors' subsequent report, the problem was not corrected until the third day of the inspection.
The inspection checklist told a different story. It said the contractor's inspector had successfully reached the hotline from several housing units on the second day.
One account put a working complaint channel inside the facility before the other account said the problem had been fixed. The difference was a day. The institutional difference was much larger: whether the government could rely on the inspection report to know what people in its custody could actually do.
The Department of Homeland Security's Office of Inspector General published that discrepancy in June 2018. Its report examined inspections conducted for Immigration and Customs Enforcement, including work by the Nakamoto Group, a private inspection contractor. The auditors had accompanied inspectors, reviewed records and interviewed officials. They were looking at the mechanism that was supposed to distinguish acceptable detention conditions from conditions requiring intervention.
The report did not establish why the contractor's checklist conflicted with the auditors' observation. It did not make a criminal finding of falsification. It did not identify a particular person whose complaint went unheard because of that discrepancy. Those limits matter. The evidence supports a narrower and still consequential conclusion: the government's own watchdog documented an assurance that did not match what its auditors said they observed.
The hotline example also explained why this was more than a dispute about paperwork. A complaint channel is part of an oversight system. If the channel fails, information may not leave the institution. If an inspection inaccurately describes the channel as functioning, the report can conceal the very obstacle that keeps outside officials from learning about other problems.
The auditors found weaknesses at later stages, too. Deficiencies could recur after correction plans were requested. People stationed inside facilities could identify problems without having authority to make managers fix them. In a second investigation, published in January 2019, the inspector general found that ICE rarely used financial penalties against facility contractors despite thousands of recorded deficiencies.
This was an institution with inspections, standards, monitors, contracts and a federal watchdog. Its opacity came partly from the distance between those layers. A problem could be visible in one office and unresolved in another. A rule could exist in a contract while an exception circulated elsewhere. A report could declare a condition acceptable without a reliable record of the work behind that conclusion.
The question was not simply whether someone had visited. It was whether the visit generated dependable knowledge, and whether that knowledge changed what happened inside.
Source: DHS OIG-18-67, June 26, 2018, pp. 6-15, particularly p. 9. Source: DHS OIG-19-18, January 29, 2019, pp. 4-15.
2. Who was inspecting whom
Immigration detention is civil custody. That describes the legal purpose of the confinement; it does not describe every person's history or resolve whether an individual should be detained. It does explain why this account should not casually turn every facility into a prison or every person held there into someone serving a criminal sentence.
The detention system examined by the inspector general included several kinds of institutions. ICE operated service processing centers, contracted directly with detention companies and used agreements involving state or local facilities. Other arrangements ran through the United States Marshals Service. The physical place, the operator and the contracting authority could therefore be different entities. Accountability depended on identifying the particular relationship rather than treating every building as the same kind of federal institution.
At the end of fiscal 2017, the 2018 report counted 211 facilities in the categories it described, holding 35,435 people. ICE's total detained population was nearly 38,000. The larger total and the table's total were not interchangeable. Family residential facilities and certain other arrangements were outside parts of the inspection review. The report's scope was substantial without encompassing every person in immigration custody.
Inside ICE, Enforcement and Removal Operations, known as ERO, was responsible for detention operations. It contracted with Nakamoto to conduct inspections. The report said that relationship began in 2007 and that the contract had been competitively awarded again in 2016. Nakamoto was the inspection company in this account, not the company operating every facility it visited.
A second inspection function belonged to ICE's Office of Detention Oversight, or ODO, located within the Office of Professional Responsibility. ODO used federal employees and assistance from another contractor, Creative Corrections. Its work was institutionally distinct from the ERO inspection contract. It was still part of ICE, rather than an outside court or a completely separate civilian regulator.
A third layer consisted of detention service managers, federal employees assigned to monitor facilities on site. In December 2017, 35 of these managers covered 54 facilities holding more than 70 percent of ICE's detained population. Their proximity could supply knowledge that a short inspection could miss. Their presence did not automatically give them command over the facility or its contractor.
Above and around those arrangements stood the DHS inspector general. Its auditors were examining the inspection system, not issuing the routine facility checklist at the center of this story. That distinction matters when two documents are described as government inspections. One was the product used to monitor conditions. The other assessed whether that product and the surrounding process could be trusted.
The reports also concerned different contracts. ICE's contract to purchase inspection services was separate from contracts or agreements to hold people in detention. A criticism of Nakamoto's inspection work does not establish that Nakamoto received the billions paid to detention operators. A penalty against a detention operator would not, by itself, establish accountability for an inaccurate inspection report.
This division of work created multiple possible points of intervention. It also created multiple opportunities for responsibility to stop at the edge of someone's assignment. An inspector could identify a deficiency. A monitor could confirm that it persisted. A contracting official could possess a remedy. None of those steps guaranteed that the next person received the necessary information or acted on it.
Source: DHS OIG-18-67, background and scope, pp. 1-5 and 14-15. Source: DHS OIG-19-18, background, pp. 1-4.
3. The size of a three-day promise
The routine inspection assignment was broad. Depending on the standards governing a particular facility, Nakamoto inspectors were expected to assess 39 to 42 detention standards containing more than 650 individual elements. The same visit could require interviews with 85 to 100 detained people, conversations with employees, daily briefings and the beginning of the written report.
The team generally had three to five inspectors and three days.
Those figures establish workload, not a mathematical proof that a competent inspection was impossible. Some records can be prepared before arrival. Some questions can be resolved quickly. Experienced inspectors may recognize a problem without repeating every step a newcomer needs. But a government purchasing an inspection must connect its requested scope to a credible account of the time, staffing and evidence needed to complete it.
The inspector general found that this connection was weak. Nakamoto inspectors themselves described difficulty covering the assignment in the available time. In the report's illustration of a five-person team, four inspectors could each handle roughly ten standards while a fifth assessed medical care. The medical assignment could involve numerous checklists and reviews of multiple records for each applicable checklist. A single heading on a completed form could therefore stand for a considerable amount of underlying work.
ODO's model supplied a useful comparison. Its teams generally had six or seven members, combining federal employees with contractor personnel. They assessed 15 or 16 core standards during a three-day visit. An individual might be responsible for two or three standards rather than approximately ten. That narrower scope gave the team more opportunity to examine the evidence behind each conclusion.
The comparison did not establish that ODO could replace Nakamoto without additional staff, money or changes in coverage. ODO visited facilities less frequently, approximately once every three years, and covered fewer than half the standards assessed in the routine inspections. A thorough examination of selected subjects leaves other subjects outside that examination. Frequency, breadth and depth were different dimensions of oversight, and each system made different tradeoffs.
The preparation differed as well. Nakamoto typically relied on its previous inspection report and a facility's self-reported information about incidents and developments. ODO gathered contracts, records, local policies and information about complaints, deaths, emergencies and uses of force. Such material could direct attention toward a question before an inspector reached the gate.
An inspection organized around the previous checklist risks inheriting the previous checklist's blind spots. That is an analytical risk, not a finding that every Nakamoto visit repeated an earlier mistake. The practical issue is whether a new inspection begins with an independent reason to investigate a subject, or primarily with the institution's own account of what has changed.
For a reader outside the facility, the completed report can make these differences disappear. Both models produce findings in official language. Both can generate a table of standards. The finished document does not necessarily communicate how many people tested its claims, which records they examined or how much time they had to resolve contradictions.
The inspector general's criticism therefore reached the procurement decision behind the visit. If the government asks for breadth without providing sufficient resources or defining acceptable methods, it helps produce the uncertainty later hidden by a confident checklist.
Source: DHS OIG-18-67, inspection preparation, staffing and scope, pp. 3-8.
4. The person answering the question
An inspection of confinement depends partly on hearing from the people confined. Their testimony cannot establish every technical fact, and an inspector should test accounts against records and other evidence. But a system that excludes or distorts those accounts loses information that administrators may not possess, may interpret differently or may have reason to minimize.
The inspector general observed Nakamoto interviews conducted in dormitories and common areas. Inspectors asked brief sets of questions in groups while facility employees were present. The auditors contrasted those exchanges with the private, individual interviews required by the inspection process. Counting the number of people in a room could make the interview total appear extensive while leaving the quality of the exchange uncertain.
Privacy has an obvious institutional purpose here. A person dependent on employees for daily needs may speak differently when those employees can hear the complaint. The report did not establish that a particular participant withheld a specific allegation out of fear. It documented an interview setting in which that possibility was not adequately controlled. A private interview is a method for reducing that risk, not a guarantee that every answer will be complete or true.
Language created another filter. During the Nakamoto visits the auditors observed, they did not see interviews conducted in languages other than English or use of the department's translation services. Inspectors selected people who could speak English. In one instance, a guard translated. The resulting conversations could not be treated as an unfiltered account of the experience of everyone held in the facility.
Translation by someone connected with custody introduces a further question about independence. Even an accurate translation does not reproduce the conditions of a confidential conversation with a neutral interpreter. The issue is structural. An auditor need not prove that a particular guard deliberately changed a sentence to recognize that the person receiving a complaint should be able to hear it without relying on the institution being examined.
ODO's observed practice was different. Its teams conducted 30 private, individual interviews, included people in segregation and used Spanish and telephone interpretation for other languages. They sought a range of ages, countries of origin, medical histories and lengths of detention. The auditors said ODO followed up on the issues raised.
That comparison identifies concrete methods, rather than an abstract preference for more sympathetic inspectors. Select people with different experiences. Speak privately. Provide interpretation. Record the issue. Check what can be checked. A government's confidence in an interview finding should depend partly on whether those steps occurred.
The inspector general's observation covered two Nakamoto inspections and two ODO inspections, conducted during its 2017 fieldwork. It was not a random survey of every interview the companies or offices had ever conducted. The article cannot turn the observed language limitation into a claim that Nakamoto never used interpreters anywhere. The evidence supports criticism of the visits examined and of the controls intended to ensure that the required methods were followed.
Within that scope, the distinction is substantial. A detention inspection that asks many people a few public questions may generate a reassuring interview count. A private conversation that produces one verifiable complaint may tell the government considerably more about whether its standards operate in practice.
Source: DHS OIG-18-67, detainee interviews and methodology, pp. 7-9 and Appendix A.
5. When an answer became evidence
The telephone discrepancy was one example of a broader concern: the difference between accepting an answer and establishing that the answer was reliable. In the inspections it observed, the inspector general found that Nakamoto sometimes treated staff assurances as sufficient even when the subject called for examining records or testing a condition.
Auditors described an inspector checking that identification documents were in immigration files without inspecting the files. They described reliance on employees' statements about drivers' licenses. They also observed two newly arrived people placed in segregation because the general population area was full, without the follow-up on documentation that the auditors expected.
Each example involved a distinct requirement, so they should not be flattened into one universal accusation. The record did not establish that every identification document was missing, that an unlicensed driver transported someone or that the segregation placement necessarily violated every applicable rule. It established shortcomings in how an inspection conclusion was reached. A claim can happen to be correct while the process used to certify it remains inadequate.
That distinction is central to inspection work. The government pays for a method that can separate a compliant condition from a noncompliant one. If the method consists mainly of asking the institution which condition exists, the government has purchased a restatement of management's position with an inspector's signature attached. Some questions can properly be answered through interviews, but the report needs to distinguish an employee's representation from independent verification.
The auditors also described questions put to employees whose responsibilities did not match the subject being assessed. A classification officer was asked about matters such as the law library or visitation. An ICE employee was asked about correctional officers' duties, while the resulting report described correctional officers as understanding the relevant standards. Knowing who supplied an answer was therefore part of knowing what the answer could establish.
The completed reports sometimes extended beyond what the auditors saw. A report represented detainees as expressing satisfaction with access to the law library, visitation and ICE officers. The auditors did not observe some of those questions being asked, and they heard detained people say they did not know who their ICE officer was. The watchdog treated the gap as a problem with the reliability of the reporting.
An observation by accompanying auditors is not omniscience. Inspectors and auditors can be in different places, and an exchange can occur outside another observer's hearing. The contractor disputed the watchdog's account, a response examined later in this article. Those considerations make the underlying notes and interview records important. They do not permit the published discrepancy to be dismissed without examining how the reported conclusion was supported.
The reasonable demand is traceability. A reader reviewing an inspection should be able to identify what was observed, what was sampled, who was interviewed and what remained unresolved. That information allows a supervisor to distinguish a verified condition from an inference and an inference from an unsupported assurance.
Without it, the report's authority depends heavily on confidence in the author. That is a weak foundation for a system whose purpose is to test someone else's assurances.
Source: DHS OIG-18-67, inspection methods and report accuracy, pp. 7-10.
6. Two counts, one caution
A comparison in the 2018 report gave the concern a numerical form. During fiscal 2016, ODO and Nakamoto inspected the same 29 facilities. ODO identified 475 deficiencies. Nakamoto identified 209, despite evaluating more standards.
The difference was large enough to require an explanation. It was not, by itself, a count of exactly 266 violations that Nakamoto had missed. The visits did not form a controlled experiment with identical dates, identical scopes and an agreed method of counting every condition. A facility could improve or deteriorate between visits. A reporting system could group findings differently. The proper conclusion begins with the divergence and then examines the methods that might explain it.
The contractor later argued that the two systems served different purposes and recorded deficiencies differently. It said some findings appeared in comments rather than checklist components that triggered corrective action. That explanation makes the design of the government's reporting form part of the accountability question.
If an observation enters a comment field without entering the correction process, its institutional effect may differ from an observation entered in a designated deficiency box. An outsider who sees only the final count may not understand the difference. A supervisor relying on that count may face the same problem unless the reporting system makes it explicit.
The inspector general's evidence did not consist solely of the totals. Its auditors had also observed differences in preparation, staffing, interviews, file review and follow-up. Those observations supplied reasons to investigate whether the broader inspection was sacrificing depth. The counts were one indicator within a larger assessment, not a substitute for that assessment.
ODO's work was not presented as flawless. The report said some ODO reports arrived late, according to ERO officials. Its narrower focus left many standards outside a given visit. Its frequency meant that substantial time could pass before a facility received another examination of that kind. A government cannot solve those limitations merely by pointing to a larger deficiency count and declaring one model sufficient for every purpose.
The more useful question is what the government learns from disagreement between its own monitoring systems. Does it compare the findings by standard, date and severity? Does it identify whether a recurring concern was absent, overlooked, corrected temporarily or recorded in a place that did not trigger action? Does it change the next inspection based on the answer?
A discrepancy can improve oversight if it initiates that work. It can also disappear into competing explanations, leaving each office confident in its own product and nobody responsible for reconciling them. The published reports did not supply a complete, condition-by-condition reconciliation for the 29 facilities.
For readers, the restraint matters as much as the number. More findings do not automatically mean more harm, and fewer findings do not automatically mean safer conditions. They can also reflect the intensity of observation and the rules for recording what is found. The government needed a way to distinguish those possibilities before using inspection results as assurances about the institutions holding people in its custody.
Source: DHS OIG-18-67, p. 7 and discussion of ODO reporting. Source: Nakamoto Group response to senators, December 4, 2018, p. 4.
7. The inspection of the inspector
A government can delegate an inspection without delegating its responsibility to decide whether the inspection is any good. That responsibility requires a second kind of examination: quality assurance over the purchased work itself. The 2018 report found that this layer was poorly documented and, in some years, absent.
The auditors could not obtain documentation of ICE quality assurance visits for fiscal 2014 or 2015. ICE acknowledged that it had conducted none in fiscal 2016. None were planned for fiscal 2017 because officials said they lacked the appropriate tools. The precise findings differ by year. Missing documentation is not proof that no visit occurred, while an acknowledged absence is stronger evidence about what was actually done.
The distinction does not make the record reassuring. A contractor producing formal judgments about detention conditions was operating without a consistently documented government process for testing those judgments. The same agency that specified the inspection assignment was unable to demonstrate regular verification of how the assignment was performed.
Quality assurance could have examined questions that later became central to the inspector general's report. Did the team have enough time? Were interviews private? Were interpreters used when needed? Did the claimed file review occur? Did the written report accurately distinguish direct observation from information supplied by facility employees? A completed checklist would not answer those questions about itself.
ODO had documented supervisory visits, which showed that another inspection function within ICE had a mechanism for observing its own work. That did not prove that every ODO conclusion was correct. It demonstrated that supervision of inspection practice was a concrete administrative function, rather than an unrealistic demand that no institution could satisfy.
Both types of routine inspection were announced in advance. Facility employees told the inspector general that advance notice could allow conditions to be changed temporarily for a visit. That testimony supports a concern about the design of the process. It does not establish that every facility staged conditions or that any particular compliant condition was necessarily artificial.
Advance notice can have legitimate uses. Inspectors may need records, access to particular employees and coordination that makes a short visit productive. Surprise can reveal ordinary conditions but can also complicate document collection. The policy question is how to combine those approaches so that logistical convenience does not become a predictable opportunity to conceal a recurring problem.
The inspector general recommended that ODO consider unannounced follow-up visits. ICE responded that fully unannounced work could be difficult because of the records needed before a visit, while describing plans for targeted follow-up and less notice. The disagreement concerned the practical design of verification, not whether a correction should ever be checked.
A reliable inspection system should be able to explain why it trusts a report without merely pointing to the existence of the report. It should also be able to distinguish a condition that persists between visits from one that appears only when visitors are expected. Those are basic questions about the product the government was buying. The watchdog found that ICE had not consistently assembled the evidence needed to answer them.
Source: DHS OIG-18-67, pp. 9-10 and management response to Recommendations 2 and 3.
8. A correction plan is a promise
Finding a deficiency begins another process. The facility must be told what needs to change. Someone must accept or reject its proposed response. Someone must determine whether the response occurred and whether it lasted. The 2018 investigation found that the movement from finding to correction was inconsistent.
The process used a document called a uniform corrective action plan. ERO field offices were required to return plans containing proposed or completed corrections to the agency's detention standards unit within 55 days. For fiscal 2015, the auditors found responses for only eight of 20 plans they reviewed. For fiscal 2016, they found responses for 12 of 25. Some responses arrived two to six months after the 55-day deadline.
Those figures concern the plans examined and the records available to the auditors. They should not be converted into a claim that every facility without a timely response left every deficiency uncorrected. A facility might make a change without producing the required document. But missing the document deprives the agency of a consistent way to establish what was promised, who was responsible and whether the promised action addressed the finding.
In some cases, the response did more than arrive late. A facility resisted the substance of the requirement. One declined to permit voluntary, free legal assistance between detained people because its own policy prohibited the practice. Another declined to put exact count times in its handbook. The inspector general described these responses as examples of a correction process that did not reliably ensure compliance.
A facility's preference does not necessarily resolve the contractual question. If the applicable detention standard requires a practice, an operator's contrary local policy creates a conflict that the contracting agency must address. It cannot be resolved simply by recording that the operator prefers its own rule. Either the requirement applies, an authorized exception changes it or the agency must decide what to do about noncompliance.
Verification was inconsistent as well. Photographs, revised policies and other evidence were not always required. Some problems were not comprehensively checked until the next inspection. The interval matters because a promise to change can be mistaken for a completed change, and a completed change on one day can be mistaken for sustained compliance over time.
The report identified repeated deficiencies involving searches, notification of alleged or proven sexual assault, handling of mail and recreation. These were not merely questions of neat filing. They concerned privacy, access to information, reporting of serious allegations and the conditions of daily confinement. The fact that different standards carried different risks made prioritization necessary, but it did not justify treating all correction plans as equivalent administrative tasks.
There were successful immediate corrections. During observed inspections, facilities stopped medical copayments, improved examination privacy, repaired telephones and updated handbooks. ODO recorded 106 on-site corrections among its 475 fiscal 2016 deficiencies. These results show why the distinction between finding and fixing should not become a claim that inspections accomplished nothing.
The problem was continuity. An inspection could produce a useful correction without the broader system establishing that every significant unresolved finding received an owner, a deadline, evidence of completion and a later check. Those are the steps that turn an inspection from an event into a means of changing institutional behavior.
Source: DHS OIG-18-67, on-site corrections and corrective action plans, pp. 8 and 11-13.
9. Inside the building, outside the chain of command
The detention service managers offered a different model of oversight. They were present on site and could encounter conditions between formal inspections. Their work included attention to food service, medical care, segregation, grievances and contact with detained people. The inspector general found that this monitoring could identify and help resolve problems.
During fiscal 2017, the managers identified 6,216 deficiencies and reported resolving 4,331. The figures show substantial activity. They do not establish that the difference between the two numbers was a verified inventory of unresolved conditions on a single common date. Findings and resolutions can occur at different points in a reporting period, and the report's aggregate should not be made to answer a more precise question than its accounting permits.
The numbers also expose a limitation of measures based only on detection. A busy monitor can generate a long list of findings because the facility has many problems, because the monitor is diligent or because both are true. Without information about seriousness, recurrence and correction, a large total is difficult to interpret as either a success or a failure of the oversight program.
The managers' authority was limited. They could identify a concern and seek support, but they could not simply direct every change needed to bring a facility into compliance. Their effectiveness depended partly on the relevant ICE field office. Some offices supported their work and involved them in local compliance teams. Other relationships were strained. The report described some monitors being regarded as a nuisance.
That description is important because it locates resistance within the government as well as within the facilities it used. A contractor's willingness to correct a problem is only part of the story. A federal employee trying to enforce the agency's standards also needs the agency's operational leadership to treat the finding as work that must be completed.
Proximity does not solve a problem of authority. A monitor can see a condition every day and still lack the power to change it. Conversely, an official with contractual authority may work elsewhere and depend on information that arrives late, incompletely or through an informal chain. Oversight requires a connection between those positions, not merely the existence of both.
The inspector general recommended protocols to support the managers and improve their collaboration with field offices. ICE agreed to work on those arrangements. This was a practical response to a practical weakness: people charged with observing compliance needed a predictable route for escalating a finding and obtaining a decision.
An effective route would also preserve disagreement. If a monitor says a problem persists and an operator says it is resolved, the system needs a record of that conflict and an accountable decision about the evidence. Otherwise a finding can disappear because one participant considers the issue closed, without a clear explanation available to the next inspector.
The managers' work complicates a simple story of total government ignorance. The agency often had information. Its own employees were identifying thousands of problems and resolving many of them. The accountability failure examined here was partly the failure to make that knowledge decisive when cooperation ended or a finding required action beyond the monitor's authority.
Source: DHS OIG-18-67, detention service managers, pp. 14-15 and Recommendation 5.
10. What the government bought after it knew
The second inspector general investigation moved from inspection practice to contractual consequences. Published in January 2019, it examined whether ICE used the tools available to hold detention facility contractors accountable for performance. This was a separate inquiry with a different population of facilities and a different period of financial review.
The investigation focused on 106 facilities for which ICE was the primary contracting authority: eight contract detention facilities, 87 facilities covered by intergovernmental service agreements and 11 dedicated facilities under that kind of agreement. It excluded ICE-owned service processing centers and facilities contracted through the Marshals Service. Its results should not be attributed to all 211 facilities described in the earlier report.
From October 1, 2015, through June 30, 2018, ICE paid the contractors operating those 106 facilities more than $3 billion. During the period, inspections and monitoring documented 14,003 deficiencies. The inspector general found only two occasions on which ICE imposed financial penalties, totaling approximately $3.9 million, or about 0.13 percent of the payments.
The comparison is striking, but it requires careful handling. A deficiency is not necessarily a separate injured person or a separate legally punishable event. The total combined findings from different monitoring activities. Repeated conditions can appear more than once, and the seriousness of individual findings varies. There is no defensible basis for multiplying 14,003 by an invented standard fine and announcing the amount the government should have recovered.
The two financial penalties also arose in different ways. One concerned repeated health and mental health deficiencies at a facility over three years. The other followed a Department of Labor finding involving underpaid wages. The latter was not a penalty imposed in response to one of the detention-standard findings highlighted in the same way as the first. The report also found no payment withholdings during the reviewed period.
Money is not the only possible consequence. Moving detained people, reducing a facility's use, demanding a correction or ending an agreement can matter more than a fine in a particular case. ICE raised that point in its response, saying it had used measures including removal of detainees, reduced usage and termination. The inspector general said ICE had not provided specific examples sufficient to evaluate those assertions within the review, and that the materials for the 106 facilities did not show contracts terminated for failure to meet detention standards.
That exchange prevents two opposite shortcuts. The small penalty total does not prove that ICE never took any action anywhere. The agency's description of other remedies does not establish that those remedies effectively addressed the documented deficiencies in the facilities and period under examination.
The underlying problem was the connection between a finding and a reasoned consequence. A government can decide that a cooperative correction is more effective than a financial sanction. It should be able to show why, what correction followed and what it did when the problem returned. A penalty count becomes meaningful when it sits within that larger record of decisions.
The inspector general found that the record and the procedures supporting it were inadequate. The agency was paying for detention while lacking a consistent account of how nonperformance changed the terms on which it continued to pay.
Source: DHS OIG-19-18, scope, financial penalties and agency response, pp. 1-9 and 17-18.
11. The missing instructions for consequences
The contracts did not all provide the same practical map for responding to failure. Only 28 of the 106 reviewed arrangements included a quality assurance surveillance plan, usually shortened to QASP. Such a plan described how performance would be monitored and how deficiencies could lead to contractual action.
All eight direct contract detention facilities had a plan. Ten of the 11 dedicated intergovernmental facilities had one. Only ten of the 87 other intergovernmental facilities did. Those categories matter because the absence was concentrated in a particular form of agreement, rather than distributed evenly across the system.
The inspector general found that the plans were the only documented source of certain instructions for issuing contract discrepancy reports and imposing penalties. Without a plan, officials did not share a consistent understanding of their options. The problem was not simply that a document with a particular title was missing. It was that the procedures needed to turn a performance problem into an enforceable response were unclear.
The auditors interviewed 11 contracting officer's representatives, employees responsible for important aspects of contract oversight. Five believed they could issue discrepancy reports without a QASP. Two believed they could not. Two believed they could issue reports but not impose financial penalties. Two did not know. That distribution described uncertainty among the people expected to put the contracting system into practice.
The report did not settle every legal question about every agreement by counting those answers. Contractual authority depends on the instrument, the applicable rules and the official's delegated role. What the interviews established was an operational weakness: personnel performing comparable oversight functions could not consistently explain the available response when a facility failed to meet a standard.
The agency also lacked a consolidated record of discrepancy reports. Information remained in individual contract files, limiting its ability to describe how often the mechanism was used, why it was used and what followed. A file may preserve a decision for one facility while failing to support an agency-wide assessment of recurring problems or inconsistent treatment.
The resulting opacity was internal as well as public. Senior officials needed to know whether comparable failures produced comparable responses, whether one field office escalated findings that another ignored and whether a contractor repeatedly avoided consequences by promising another correction. Without a usable record, answering those questions required reconstructing events across separate files.
Standardization alone would not guarantee good judgment. A rigid system that imposed the same consequence for every defect could be unfair and ineffective. The inspector general's recommendation allowed for documented reasons when a surveillance plan could not be included and called for clarification of the actions available under the remaining arrangements. The aim was an intelligible decision process, not automatic punishment detached from context.
The public interest is straightforward. A detention contractor should know what performance the government requires. The government employee monitoring that performance should know how to respond when it fails. An official deciding against a penalty should leave enough evidence to explain the choice. When those conditions are missing, the strongest rule in a contract can become uncertain at the moment someone needs to enforce it.
12. When the rule acquired an exception
A detention standard can be changed in practical effect without being rewritten for every facility. An authorized waiver can permit a particular facility to depart from a requirement. That can be a legitimate administrative tool when a rule cannot sensibly be applied in its ordinary form or an alternative provides an acceptable safeguard.
It can also make the published standard an incomplete guide to what the government actually permits.
The 2019 report reviewed 68 waiver requests submitted between September 2016 and July 2018. ICE approved 65, approximately 96 percent. Only three approved waivers included an expiration date. The remaining 62 did not. Those numbers describe a set of requests and approvals; they do not prove that every exception was improper or that every request deserved rejection.
The inspector general's concern was the process surrounding the exceptions. ICE lacked formal policies defining parts of the waiver procedure, officials' authority was unclear, and waivers were not consistently distributed to all the people responsible for monitoring performance. An exception could alter the practical meaning of a requirement without becoming reliably available to the officials charged with enforcing it.
One approved waiver permitted mixing people assigned to different custody levels. Another permitted a chemical agent that differed from the one ordinarily authorized. These were decisions with implications for how a facility operated, not merely changes to the appearance of a form. Evaluating their merits required an explanation of the risk, the alternative and the authority for accepting it.
A waiver without an expiration date is not necessarily perpetual permission in every legal sense. It might be reviewed or withdrawn through another process. But when the system does not require a clear review point, an exception granted for a temporary problem can survive without a documented decision that the original justification still exists. The absence of an end date makes the review mechanism more important, not less.
There was also a contracting question. Federal procurement rules reserve contract modification authority to contracting officers in the settings where those rules apply. ICE disputed aspects of the inspector general's application of the Federal Acquisition Regulation to intergovernmental agreements, distinguishing them from direct contracts. The watchdog also cited ICE guidance calling for use of procurement principles to the extent possible in those arrangements.
The article cannot turn that disagreement into a judicial finding that every waiver was illegal. The inspector general identified a risk that waivers could circumvent contract terms or be approved without demonstrated authority. Resolving a particular case would require the agreement, the waiver, the delegation of authority and any modification that followed.
The information problem was independently serious. Acquisitions personnel told auditors they were not always aware of waivers affecting the contracts. Monitors could therefore be assessing compliance against requirements that another part of the agency had excused, while contracting officials might lack the record needed to determine whether the exception required a formal change.
An accountable exception should be easy to follow: what requirement changed, why it changed, who authorized the change, what safeguards replaced it and when the decision would be reconsidered. The inspector general found that ICE had not consistently built that chain. The result was a system in which knowing the published rule did not necessarily tell an observer which rule governed the facility.
Source: DHS OIG-19-18, waivers, pp. 9-12 and Recommendation 3 and responses, pp. 19-20.
13. The official whose supervisor needed the beds
Contract oversight is a job performed by people inside an organization. Their authority depends on formal delegations, but their daily decisions also depend on workload, access, travel and the supervisors who evaluate their performance. The 2019 investigation examined those conditions because a contractual remedy is useful only if someone can credibly initiate it.
ICE had previously placed contracting officer's representatives under headquarters supervision, with input from field offices. In May 2010, it moved their direct supervision to the field. The stated explanation was that the earlier arrangement had become cumbersome and suboptimal. Officials interviewed by the inspector general did not provide an additional explanation for the change.
Field supervision can offer practical advantages. Local managers know operational demands and can coordinate employees working in the same region. The concern arises when an employee responsible for challenging a contractor reports to an official whose operational work depends on that contractor's continued availability. The arrangement can put contract enforcement and immediate operational convenience in the same supervisory relationship.
Two representatives told the auditors they hesitated to identify noncompliance or issue discrepancy reports because they feared retaliation from field management. That was evidence of expressed concern, not a finding that retaliation had occurred. Other contracting officials worried that representatives might be reluctant to disagree with the supervisors responsible for their appraisals. The report treated the structure as a threat to independent oversight.
One example involved a representative authorizing payment for transportation at a field supervisor's request even though the contract did not provide for transportation. According to the report, the representative lacked authority to make that contract change. The episode illustrated how an operational request could cross the boundary between administering an agreement and altering what the government had agreed to purchase.
Work assignments created another pressure. Representatives could be tasked with duties outside detention contract oversight, including background investigations, supervision of support personnel and work on unrelated office requirements. These assignments were possible under job descriptions allowing other duties. The report did not say that every additional task was unnecessary; it questioned the effect on the employee's central oversight function.
Some employees also carried extensive contract portfolios. Of the 11 representatives interviewed, four oversaw ten or more facilities. Two had responsibilities involving 16 and 22 contracts across multiple states. Five also oversaw major transportation contracts. One was project manager for a processing center intended for four federal agencies while overseeing three detention contracts.
Travel restrictions added a practical barrier. Three field offices restricted representatives from traveling to detention facilities, limiting their ability to evaluate performance directly. Other offices provided administrative support or allowed networks of assistants. Those differences showed that the formal job title did not guarantee the same capacity to perform the job in every location.
The inspector general recommended a review of workload, staffing and organizational placement. ICE agreed to consider those issues, including whether representatives should remain under field supervision. This was not a promise that every reporting line would change. It was recognition that enforcement depends on the conditions under which the enforcing employee works.
An institution can assign responsibility without supplying independence. It can also supply independence on paper while leaving an employee too overloaded to use it. The report's personnel findings explain why accountability cannot be measured only by counting available contractual remedies. Someone must have the time, access and organizational support to choose one when the choice is inconvenient.
14. The contract the monitor could not readily see
One of the least dramatic findings in the 2019 report may be among the most revealing. Employees charged with monitoring detention contracts did not consistently have access to the complete contracts and modifications they were expected to oversee.
ICE's acquisitions office maintained documents in an electronic system called PRISM. Some contract management personnel had access. Contracting officer's representatives and detention service managers did not. Documents were also placed on a shared drive intended to become the official repository, but the report said those monitoring employees lacked access there as well.
Older agreements presented another obstacle. Some active contracts originating around 2005 or earlier had historical paper files. The monitoring employees did not automatically receive access to those materials. A complete understanding of the government's obligations could therefore depend on obtaining documents from other personnel rather than consulting a reliable, current record directly.
Representatives were supposed to maintain their own files, including the agreements and modifications they oversaw. That expectation did not solve the distribution problem. One interviewee said contracting officers or specialists should provide the documents, but the practice was inconsistent. Another described assembling her own electronic collection when she began the job.
There is nothing inherently improper about keeping a working copy. The risk is divergence. If an employee's file lacks a modification, the employee may assess a contractor against an outdated requirement. If another employee has a waiver that never reaches the contract file, the two may disagree about what compliance means without realizing they are consulting different records.
The same problem affects institutional memory. A diligent employee can reconstruct a file through individual effort. When that employee changes jobs, the institution needs a durable record that the successor can understand. A system dependent on personal collections is vulnerable to ordinary turnover, even without any attempt to hide information.
The inspector general recommended efficient access to the complete contract record. ICE responded that it required electronic availability of contract documents and would give the relevant monitoring personnel read-only access. The proposed fix was modest in form. Its importance lay in connecting the people who observed performance with the documents defining that performance.
This is a particular kind of opacity. It does not require a secret law or a locked public archive. The rule may exist in a government file while remaining difficult for the responsible employee to retrieve. The institution can then appear heavily documented from the outside and still be poorly informed at the point where a decision must be made.
For a detained person, the distinction among a contract, a modification, a standard and a waiver may be invisible. The immediate experience is whether medical care is available, a grievance is handled or a required opportunity is provided. For the government employee responsible for enforcing the agreement, those documents determine what action can be demanded and who can demand it.
A usable record cannot guarantee that an official will act. It can remove an avoidable excuse for inaction and make the decision reviewable afterward. That is why access to the complete agreement belongs in the same accountability story as inspection quality and financial penalties. The enforcement chain is only as intelligible as the documents connecting its participants.
Source: DHS OIG-19-18, access to contract files, pp. 14-15 and Recommendation 5 response, p. 20.
15. What the contractor and the agency said
Nakamoto contested the watchdog's account. In a December 4, 2018 letter to senators, the company denied inaccurate reporting and said confidential interviews and relevant staff interviews had occurred. It emphasized inspectors' experience, described annual inspections as snapshots and said ICE supplied the reporting framework and approved its work. It also said it had formalized private interviews. These are the company's assertions; this article has not independently verified the underlying interview records.
That response deserves substantive treatment. An inspector cannot observe every condition on every day. An annual visit should not be represented as a guarantee about all intervening months. A reporting form designed by the government can constrain how findings appear, and the government bears responsibility for the product it requests and accepts.
Those points do not fully answer the watchdog's central criticism. Some findings concerned what the auditors said they observed during the same visits they accompanied, including the hotline discrepancy and interview practices. Explaining differences between separate inspection dates does not resolve a contradiction about the work performed during an observed inspection. Resolving that contradiction would require records specific to the disputed claim.
The contractor's explanation of different counting methods also directs attention to the connection between documentation and correction. If the government receives significant information in a narrative comment, its process should explain whether that information requires action. A concern should not lose practical significance merely because it appears outside the field used to calculate a headline total.
ICE's formal responses were broader. In the 2018 report, the agency agreed with all five recommendations. It proposed revisiting the inspection statement of work, strengthening quality assurance, conducting targeted follow-up, improving verification of corrective actions and supporting detention service managers. Those commitments addressed the architecture of the process rather than only defending individual checklists.
In the 2019 contracting report, ICE again concurred with all five recommendations while disputing parts of the findings. It defended its layered monitoring system and described nonfinancial actions available against facilities. It also said it already maintained and distributed waivers. The inspector general responded that distribution on request did not ensure the relevant employees knew the exceptions existed and could use them in their work.
The promised contracting changes included clearer guidance on discrepancy reports and penalties, tracking of those actions, a formal waiver process with recurring review, attention to oversight staffing and better access to documents. These were not trivial concessions. They corresponded to the points at which the investigators found that information failed to produce a consistent institutional response.
At publication, the recommendations were classified as resolved and open. In this context, resolved meant that the proposed actions were responsive enough to address the recommendation if completed satisfactorily. Open meant the inspector general still required evidence before closure. Neither word, read alone, establishes that the problem had already been fixed.
This article does not treat those publication-time classifications as the recommendations' status in September 2026. The reviewed record does not include a complete later closure history or the evidence underlying each eventual status change. It therefore supports an account of what the watchdog found and what ICE promised, rather than a claim that every identified weakness persists today.
That limit leaves a concrete accountability question: what evidence demonstrated that the promised changes altered inspection practice and contract enforcement? A new procedure can be necessary. The final test is whether employees used it and whether the resulting decisions can be traced from finding to outcome.
Source: Nakamoto Group response to senators, December 4, 2018, pp. 1 and 3-5. Source: DHS OIG-18-67, management response and OIG analysis, pp. 16-19. Source: DHS OIG-19-18, management response and OIG analysis, pp. 17-21.
16. A finding needs a destination
The two investigations describe a sequence of decisions that can be examined separately. A standard defines an obligation. An inspection tests it. A report records the result. A correction process assigns work. A monitor checks the outcome. A contracting official decides what continued failure means for the agreement. A later reviewer determines whether the system performed as intended.
The sequence can fail without every participant failing. An inspector can find a real problem and a facility can fix it promptly. A monitor can work diligently while lacking authority over a resistant operator. A contracting official can possess a remedy but lack a clear record or a reliable account of the applicable requirement. Describing those distinctions is more useful than assigning one undifferentiated judgment to everyone involved.
The reports also show why the volume of oversight is an incomplete measure. More visits do not necessarily produce better information if the methods are weak. More findings do not necessarily produce better conditions if the findings lack a route to correction. More written rules do not necessarily create a clearer obligation if exceptions are issued without consistent authority, distribution or review.
A stronger public account would connect those measures. For each significant deficiency, it would identify the applicable standard, the evidence supporting the finding, the responsible official, the promised correction and the verification result. If the government chose a nonfinancial remedy, the record would explain the choice and its outcome. If it approved a waiver, the record would identify what changed and when the exception would be reconsidered.
Such an account would need appropriate protection for personal and medical information. Public accountability does not require publishing a detained person's confidential file. It requires enough information about government decisions to assess whether the relevant obligation was enforced. Aggregate data can support that assessment, but only if the definitions and denominators remain clear and specific cases can be traced without exposing protected details.
The missing later closure record is also a reporting agenda. To establish what happened after the promises, a reporter would seek revised inspection requirements, quality assurance observations, completed corrective action verifications, waiver registers, discrepancy reports and the evidence used to close recommendations. Those documents could show meaningful improvement. They could also show that a formal change left the practical weakness intact. The answer should follow the records rather than precede them.
The evidence already reviewed supports a definite historical conclusion. Federal auditors found that parts of ICE's detention oversight system did not reliably establish what conditions existed or ensure that documented deficiencies led to sustained correction. The agency accepted recommendations to change the system. The contractor disputed important parts of the auditors' account. The resulting record is both substantial and bounded.
The telephone remains a useful way to understand the stakes. A complaint channel was supposed to connect a person inside with an official outside. According to the watchdog, it did not work when tested. A separate report placed successful access earlier than the auditors said the correction occurred. To a reader seeing only that report, the safeguard could appear to exist at the moment it was needed.
An accountable institution must be able to get beyond that appearance. It must show how it knows a safeguard works, what it does when the safeguard fails and who verifies the answer. The distance between those questions and a checked box is where this story takes place.
Source: Institutional analysis based on DHS OIG-18-67 and its recommendations. Source: Institutional analysis based on DHS OIG-19-18 and its recommendations.
Update and correction history
No substantive post-publication changes recorded.