Bureaucracy Times

All the news fit to be declassified.

Category: Institutional Bureaucracy

  • The Furnace Broke. The Recommendations Didn’t.

    The VA Office of Inspector General visited the Thomas E. Creek VA Medical Center in Amarillo for four days in February 2026 and found a hospital that had spent the previous three months heating itself with the equipment version of a séance. The resulting Healthcare Facility Inspection report reads like most of its siblings in this OIG series, calm, procedural, and quietly damning. Two recommendations. Zero questioned costs. A director who concurred with everything and asked, politely, for one of them to be closed before the ink dried. It is the kind of report that could put you to sleep if you didn’t notice that the sleeping happened in a hospital where nobody was checking whether the medication was still good.

    A Building That Kept Its Own Time

    In late November 2025, the heating system at Amarillo’s main hospital stopped working. Not a fuse, not a thermostat, an aged air handling unit whose replacement parts, executive leaders told inspectors, “were not readily available to order.” The outage ran for approximately three months. Engineering staff got by reviving old steam furnaces that had presumably been left in place for exactly this occasion, and when that wasn’t enough, they deployed space heaters and moved veterans out of a medical surgical unit that never did get adequate heat back. The Associate Director said the facility’s infrastructure plan includes replacing other aging systems “to prevent similar disruptions in the future,” which is the sentence every institution says right after the thing it was supposed to prevent has already happened.

    This is presented in the report’s Culture section, under the heading “System Shocks,” as a story about resilience. The staff adapted. The mission continued. Nobody froze. All of that is true, and none of it changes the fact that a facility with a $597 million medical care budget kept a wing of veterans warm with the hospital equivalent of camping gear for a full winter because a part could not be sourced. The OIG made no recommendation on this. It did not need to. The next section of the report supplies its own commentary.

    The Room Where Nobody Checked the Weather

    The Environment of Care section is where the actual grading happens, and Amarillo did not do well. Inspectors touring the medical surgical units, the ICU, the emergency department, and the community living center (VA’s term for its nursing homes) found wall damage in the emergency department and CLC hallways, damaged floors in the CLC, stained ceiling tiles over the emergency waiting area and the CLC dining room, and, in the CLC’s clean utility room, dirty equipment and items simply left on the floor next to whatever was supposed to stay clean. VHA Directive 1608(1) requires facilities to keep those two categories physically separated for the obvious reason that “clean” and “next to a mop bucket” are not the same condition.

    Then there’s the medication rooms. Staff in the ICU and the CLC were not monitoring temperature and humidity, a basic requirement under VHA Directive 1108.07(2) that exists because medications stored outside 68 to 77 degrees Fahrenheit and 30 to 60 percent relative humidity can degrade without anyone noticing. The Associate Director told inspectors he had no idea staff weren’t doing this. The facility’s own written policy, in place since 2023, said they were supposed to be. Nobody had checked whether the policy was being followed until the OIG did.

    The facility’s fix for this one is worth sitting with. In its written response, Amarillo reported that nursing staff, when interviewed, “were unable to speak to the established process,” which is a generous way of saying the people responsible for the room didn’t know there was a room to be responsible for. The corrective action was staff education and new signage identifying a point of contact. The facility then reviewed a year of monitoring data, found humidity excursions had in fact occurred, decided none of the affected medications needed to be thrown out, and asked the OIG to close the recommendation. The OIG agreed, on September 28, 2026, the same day this report was published. Recommendation open, evaluated, and closed inside the same document. That is not oversight moving quickly. That is oversight arriving exactly in time to watch the facility grade its own test.

    A Finding With Its Own Frequent Flyer Program

    None of this would be especially newsworthy if it were unique to Amarillo. It isn’t. The VA Indiana Healthcare System’s inspection, completed earlier in this same cycle, cited supplies and debris left on a clean supply room floor and chipped, damaged walls with visible residue in a patient room, language close enough to Amarillo’s that it could be a find and replace job. The VA Western Colorado Healthcare System’s inspection went further, issuing eight recommendations that included the identical clean and dirty item separation failure, plus ceiling vent dust, wall repair, and a call for the regional VISN office itself to start actually overseeing the environment of care program it was supposed to be overseeing all along.

    Three different facilities, three different states, three different inspection windows, and the same finding under the same directive with the same prescribed remedy: develop a plan, educate staff, monitor compliance until it holds for six consecutive months. VHA Directive 1608(1) has been in place since 2021. It has been amended once, in 2023, seemingly without anyone updating the muscle memory of the maintenance staff tasked with following it. This is not three unlucky facilities. This is what a compliance program looks like when the response to a finding is standardized but the underlying behavior isn’t.

    Closed, In the Sense That the Paperwork Says So

    The Amarillo report closes on an uneasy note of institutional self-awareness. In its conclusion, the OIG mentions, almost in passing, that it is tracking VHA’s ongoing restructuring, the collapse of 18 regional networks down to five, completed just months before this inspection was published, and that the agency plans to “monitor VHA’s change management” going forward. It is a tidy sentence to end on, and it says nothing about who, in a newly consolidated five-network structure, will be responsible for making sure the next facility’s clean utility room stays clean.

    The honest reading of this report is not that Amarillo is a badly run hospital. Employee survey responses describe leaders who communicate often, veterans who travel from three states away specifically to be treated there, and a 94 percent trust rating that beats the VA average. The honest reading is that a facility can do nearly everything right and still leave a heating system to fail for three months because a part wasn’t in stock, and still let a medication room go unmonitored for however long it takes an inspector with a clipboard to notice. The OIG will be back in roughly three years to check again. Given the pattern at Indiana and Western Colorado, there is no particular reason to expect the sequel to look different, only that it will, once again, be resolved with an action plan, a monitoring period, and a director who concurs.

    Fediverse Reactions

  • The VA Hospital That Kept Failing the Same Inspection

    The VA Office of Inspector General spent four days in August 2025 walking the halls of the VA New Jersey Healthcare System, and what it found reads less like a discovery than a rerun. Untested fire doors. Exposed wiring. Expired hand sanitizer. Missing sprinkler head plates. Every one of these had already been flagged, by two other auditors, in two prior years. The OIG simply confirmed that nobody had done anything about it (VA OIG, Report 25-00242-195, Sept. 25, 2026).

    A Facility With Three Report Cards and One Grade

    The Joint Commission inspected East Orange and Lyons in the spring of 2024 and found damaged ceiling tiles, emergency pull cords wrapped uselessly around handrails, and a hand sanitizer dispenser bolted directly above an electrical outlet, an arrangement that National Fire Protection Association code exists specifically to prevent. Weeks later, the Veterans Integrated Service Network’s own annual safety evaluation found the same wiring problems, the same missing sprinkler plates, the same neglected eye wash stations. Then in August 2025, sixteen months and one more VISN evaluation later, OIG inspectors arrived and cataloged the identical list. Bugs in light fixtures. An eye wash station blocked by trash cans. Another behind a locked door. A blanket warmer with a broken thermometer, so nobody could confirm it wasn’t running hot enough to injure a patient.

    The most damning line in the entire report is not a violation; it’s an explanation. Facility leaders told the OIG that when a Joint Commission finding came in, they assigned a “champion” to fix that specific item, track it for a few months, and close it out. Nobody then checked whether the same defect existed two floors up, or in the other building, or in the room next door. The system was built to close tickets, not to fix conditions. That is not an oversight. That is a workflow, and it is the reason three separate inspections over two years kept finding the same broken country.

    The Clock That Stopped in 2018

    Buried in the fire safety section is a detail that deserves more alarm than it got. Components of East Orange’s sprinkler system were last fully tested in 2018. Federal fire code requires that testing every three years. Nobody inside the facility caught the seven year lapse. It took an outside inspector, flipping through maintenance logs during a routine three year cycle visit, to notice that the system meant to put out fires in a building full of bedridden patients hadn’t been checked since the first Trump administration.

    The facility’s written response is a masterclass in bureaucratic redemption arc, promising a “comprehensive action plan,” an assessment of all six hundred fire doors, “interim life safety measures,” and monthly reporting up a new committee chain. All of this is true and all of it is also what gets written every time an inspector shows up with a clipboard. The test is not the pledge; it’s whether the OIG’s next visit, on its own three year cycle, finds anything different.

    Nobody Told the Family

    The report’s second major thread involves adverse events, VHA’s term for incidents that harm or nearly harm a patient. In the twelve months before the inspection, the facility recorded a delayed diagnosis, a serious injury, and a death, each the kind of event that VHA policy requires be formally disclosed to the patient or their family. The facility disclosed none of them. Risk managers told OIG inspectors they had “varying interpretations” of which adverse events actually triggered the disclosure requirement, an admission that a policy meant to protect patients had quietly become optional at the point of use.

    Separately, patient safety staff had let twenty nine corrective actions from prior root cause analyses go overdue, and executive leadership said they had no idea. The chief of quality had been on the job for fourteen days when the OIG walked in and told her about a backlog that had been accumulating on someone else’s watch for months.

    The Corrective Action Plan Is Also a Genre

    None of this is new to VA, and none of it is unique to New Jersey. The pattern here, an inspection generates a promise, the promise generates a committee, the committee generates a monthly report, and the underlying condition waits patiently to be rediscovered by the next auditor, is close to the operating system of federal accountability itself. VA New Jersey’s leadership turned over almost entirely in the year before the inspection, a new director, a new chief of engineering, a new quality chief, each inheriting a facility mid collapse and each producing, on cue, a fresh action plan with a fresh target date of March or December 2027.

    Reading eleven recommendations and eleven corresponding concurrences, each attached to a specific committee that will “monitor” compliance until some threshold of ninety percent is reached for six consecutive months, one starts to suspect that the compliance theater has become more sophisticated than the compliance itself. The VISN director concurred in one paragraph. The facility director concurred in one paragraph. Both thanked the OIG for the opportunity.

    The veterans using that emergency department, the ones navigating a locked interior route or an outdoor ramp to reach it, are unlikely to be reassured by a committee’s monitoring cadence. They would probably just like the doors to close.

    Fediverse Reactions

  • CMS Found Possible Errors in Medicaid Financial Data. Then It Didn’t Follow Up.

    The Centers for Medicare & Medicaid Services had a system for checking whether states submitted complete Medicaid managed-care financial reports.

    It also had a system for checking whether some of the numbers in those reports might be wrong.

    What it apparently did not have was a procedure for what to do when the second system found something.

    A new Department of Health and Human Services Office of Inspector General report found limitations in CMS’s process for verifying state-submitted medical loss ratio data, including dozens of reports containing potential inaccuracies that CMS identified but did not pursue with the states involved.

    The findings concern Medicaid managed-care medical loss ratios, or MLRs. An MLR measures the percentage of a plan’s revenue spent on covered health care services and quality-improvement activities. The federal standard is 85 percent, meaning a plan reaching that figure spent 85 percent on those purposes and the remaining 15 percent on administration and profit.

    The numbers aren’t merely informational. States use past MLRs when developing future capitation rates, the fixed monthly amounts paid to managed-care plans for each enrolled beneficiary. States are required to develop those rates so plans can reasonably achieve an MLR of at least 85 percent.

    That makes the accuracy of the underlying data rather important.

    CMS Couldn’t Check More Than Half the Reports

    OIG examined CMS’s oversight of 595 final MLR summary reports submitted by 44 states for the 2019 reporting year.

    CMS used five accuracy checks covering four required MLR data elements. According to OIG, those checks identified 57 reports submitted by 16 states containing at least one potential inaccuracy.

    There was another problem.

    For three of the four required data elements, CMS needed additional information to perform its accuracy checks. But those underlying data elements were optional for states to submit.

    The predictable happened.

    CMS could not conduct each of four accuracy checks on more than half of the 595 reports. Depending on the check, CMS could test only 43 to 46 percent of them.

    Where CMS did have enough information to perform the checks, it found potential problems.

    Among the reports CMS could test, it identified 23 potentially inaccurate MLR numerators, 15 potentially inaccurate denominators using one check, 18 using another denominator check, and one potentially inaccurate adjusted MLR. OIG noted that additional optional data might have allowed CMS to identify additional potential inaccuracies.

    CMS performed another check by comparing reported “member months” against enrollment information. It could perform that comparison for 448 of the 595 reports and identified 25 with potentially inaccurate member-month figures.

    These were potential inaccuracies, not OIG findings that the figures were definitively wrong.

    Determining that would have required another step.

    And Then, Nothing

    CMS itself told OIG that potential inaccuracies require further evaluation and possibly follow-up with states to determine whether the numbers are actually incorrect.

    CMS did not follow up with the 16 states.

    OIG found that the agency therefore missed an opportunity both to obtain corrected information where errors actually existed and to alert states so they could investigate the questionable data themselves.

    Curiously, CMS did have procedures for chasing down missing information.

    For the same 2019 reporting year, CMS followed up with states over missing MLR data and successfully obtained 192 initial or resubmitted reports from 19 states.

    Its written procedures simply did not specify what employees should do about potential inaccuracies detected by CMS’s own accuracy checks.

    Completeness, apparently, had a workflow. Accuracy was still waiting for one.

    The Findings Stayed in Their Lane

    OIG found another disconnect inside CMS.

    As of October 2025, the CMS staff reviewing state-submitted MLR data did not share the results of their accuracy checks with the CMS employees responsible for reviewing states’ future capitation rates.

    Those rates are precisely where historical MLR information can matter.

    CMS reviews the prospective rates and approves managed-care contracts containing the final rates. OIG concluded that by not communicating the accuracy-check results internally, CMS missed an opportunity for those findings to inform its rate reviews.

    OIG recommended that CMS establish procedures for handling potentially inaccurate data, share accuracy-review information with employees reviewing capitation rates, and improve the effectiveness of its MLR oversight, potentially by adding automated accuracy checks to its reporting portal.

    CMS concurred with all three recommendations.

    An Important Date on the Report Card

    There is a significant caveat to the findings.

    Although OIG issued the report in September 2026, the data it analyzed came from the 2019 reporting year.

    CMS emphasized that point in its response, calling 2019 one of the earlier reporting periods for state MLR submissions and saying it has made substantial improvements since then. Those changes include an MLR toolkit, an optional standardized reporting template, a web-based submission portal and public release of plan-specific MLR data.

    The agency also said it has already added automated checks preventing states from submitting reports with blank required fields and will consider expanding the system to perform automated accuracy checks. CMS has encouraged states to submit the optional data needed for additional verification and says it intends to continue doing so.

    That means the OIG report does not establish that the same deficiencies persist today.

    It does establish something narrower, and peculiar enough on its own.

    CMS had financial reports that mattered to Medicaid managed-care oversight. It built checks designed to identify questionable numbers in those reports. Those checks found potential inaccuracies in reports from 16 states.

    And when the warning lights came on, there was no procedure telling the agency what to do next.

    Fediverse Reactions

  • DOT Systems Missed Required Privacy Reviews as Staffing and Funding Gaps Undercut Compliance

    Three years after a data breach exposed personal information belonging to roughly 237,000 current and former federal employees, the Department of Transportation was still failing to consistently perform required annual privacy assessments on information systems containing personally identifiable information, according to a new inspector general audit.

    The explanations were considerably less sophisticated than the technology involved.

    At one DOT agency, officials could not fully explain why required privacy work had not been completed because the employees previously responsible for the program had left. At another, privacy responsibilities were divided among offices and officials acknowledged that completing the work had not been made a funded priority.

    The findings come from a Department of Transportation Office of Inspector General audit examining whether DOT had established an effective privacy program and implemented required controls for protecting personally identifiable information, or PII.

    OIG identified 247 DOT systems containing PII and selected a statistical sample of 55 for review. Nine of those 55 systems — roughly 16 percent of the sample — did not receive required annual privacy risk assessments.

    The audit does not establish that those systems were breached or that information contained within them was improperly accessed. Rather, the missing assessments meant DOT could not demonstrate that it was consistently performing one of the controls intended to identify and address changing privacy risks.

    That distinction matters. So does the history.

    In May 2023, DOT disclosed a breach involving its Federal Transit Benefit Program that exposed PII belonging to approximately 237,000 current and former federal employees. OIG had also previously identified weaknesses in DOT’s implementation of privacy policies in audits issued in 2018 and 2019.

    The latest audit suggests that some of the basic compliance problems persisted.

    At the Federal Motor Carrier Safety Administration, OIG found that three systems did not receive required annual privacy assessments during fiscal years 2022 through 2024: SAFETYNET, the Enforcement Management Information System, and the National Consumer Complaint Database.

    The National Consumer Complaint Database also lacked required Privacy Threshold Analyses during the period reviewed.

    When auditors sought an explanation, FMCSA officials could not provide documentation or direct knowledge explaining why the assessments had not been completed.

    There was a bureaucratic complication: according to OIG, the personnel responsible for managing FMCSA’s privacy program from fiscal years 2019 through 2024 had left the agency.

    In other words, the institutional knowledge apparently departed with the employees.

    The Federal Aviation Administration had its own version of the problem.

    OIG found six FAA systems that had not received required annual privacy assessments during fiscal years 2022 through 2024. Another system received an assessment in fiscal 2023 but not in fiscal 2024.

    The affected systems identified by OIG included the National Flight Data Center, National Airspace System Resource System, Flight Activity and Crew Tracking System, Weather Systems Processor and Weather Camera System.

    FAA officials attributed the deficiencies partly to how privacy responsibilities were distributed between privacy personnel and individual system owners.

    But OIG also reported a more straightforward explanation.

    According to FAA’s chief privacy officer, Air Traffic Organization leadership had not made the required privacy work a funded priority, limiting the organization’s ability to complete necessary privacy documentation.

    That left a federal privacy requirement competing for attention and resources inside an agency responsible for some of the country’s most consequential transportation systems.

    OIG identified a broader management problem as well.

    During the audit period, DOT lacked a formally approved department-wide privacy program plan meeting federal requirements. Department officials provided auditors with various policies, procedures, strategic documents and other materials that they maintained collectively constituted such a program.

    OIG disagreed.

    Auditors found DOT could not demonstrate that its senior agency official for privacy had formally approved or designated those materials as the Department’s privacy program plan. Some strategic documents were unsigned and lacked required goals and objectives.

    DOT has since corrected that deficiency.

    In January 2025 — after the fiscal 2024 period examined by auditors — the Department completed and formalized its privacy program plan. That makes the audit less a story about DOT currently operating without a privacy plan than one about how inconsistently required privacy controls were implemented before and during the period under review.

    OIG ultimately issued just two recommendations.

    The inspector general recommended that DOT maintain its newly established enterprise privacy plan and ensure that annual privacy risk assessments are performed for Department information systems.

    DOT concurred with both recommendations. OIG considers them resolved but open until the Department completes the promised corrective actions.

    The audit therefore stops short of uncovering another major DOT data breach.

    Instead, it documents something less dramatic but arguably more familiar: federal agencies had established privacy requirements, officials knew assessments were supposed to be performed, and some of those assessments nevertheless went undone because personnel changed, responsibilities were fragmented and resources were directed elsewhere.

    For a Department still operating in the shadow of a breach affecting 237,000 people, those are precisely the mundane administrative failures privacy controls are supposed to prevent from becoming something considerably less mundane.

    Fediverse Reactions

  • An Electric Griddle Cost $54 Million, According to the Federal Government

    The Federal Bureau of Prisons wants Congress to believe it has a plan for the five billion dollars it just received, three billion of which is earmarked for hiring and retaining staff under Public Law 119-21. To have a plan, an agency needs to know what it owns, what that equipment costs to keep running, and what it will cost to replace. The BOP has a system for tracking exactly that. It is called the Total Maintenance System, and according to a new DOJ Inspector General audit, it once recorded that a single electric griddle, placed into service in 2017, had racked up more than fifty four million dollars in maintenance costs. The actual figure was about four thousand dollars, a rounding error’s rounding error, off by a factor of roughly thirteen thousand. (See “Audit of the Federal Bureau of Prisons’ Acquisition and Life Cycle Management of Major Equipment Supporting Food Services,” Report 26-100, September 2026.)

    BOP officials told auditors this was a manual input error, which is technically true in the way that every clerical mistake is technically true. The more interesting question is why nobody caught it. The griddle sat in the system, presumably serene in its fictional status as one of the more expensive appliances in the federal government, until an outside auditor happened to ask.

    The System of Record That Nobody Trusted Enough to Use Correctly

    TMS exists to give BOP leadership a reliable inventory of the more than 5,700 pieces of major food service equipment spread across 118 institutions, so the agency can plan replacements and budget accordingly. When a new asset is entered, the system automatically builds a maintenance schedule and lets staff log work orders, labor, and material costs against it. In principle this is exactly the kind of system that should make a five billion dollar hiring and infrastructure infusion easier to spend well.

    In practice, auditors compared TMS records against physical equipment at two institutions and found the data barely resembles reality. At FMC Lexington, the system listed 58 major food service items; fourteen more were in actual use but never entered at all, meaning no maintenance plan or cost tracking existed for them. Of the items that were listed, the assigned condition codes were wrong across the board; eighteen items were coded as “obsolete or inoperable” when only one actually was, while items genuinely falling apart were logged as merely fair or poor. At FCC Butner, auditors found the reverse problem in miniature; eighteen items on the books could not be located at all, and six more had been quietly replaced or retired without anyone updating the record. Officials blamed manual data entry, thin quality assurance, and a general lack of training on how to use the system, which is a polite way of describing an inventory that functions mostly as folklore.

    None of this is abstract. The BOP is currently paying to rent 168 mobile refrigeration units and off-site cold storage across 52 institutions, at a cost of more than nine million dollars over six years, because its own refrigeration units are too often broken to trust with food that inmates are legally entitled to eat safely. At FCC Butner alone, six years of rental and off-site storage cost about $1.93 million, roughly $729,000 more than simply buying new freezer storage would have cost. That is not a rounding error. That is a spreadsheet nobody opened.

    A Definition Problem Dressed Up as a Data Problem

    The deeper issue, which the griddle merely dramatizes, is that the BOP has never settled on what “major food service equipment” means. A 2012 Central Office memo defines it one way, the 2023 Facilities Operations Manual defines it another way, and institution-level staff apply a third, broader definition that includes ice machines, mixers, and exhaust hoods nobody at headquarters considers major at all. Since the entire replacement-planning apparatus, TMS included, depends on knowing which assets even count, an agency that cannot define its own equipment cannot be expected to track its cost with any precision. The griddle’s fifty four million dollar phantom debt is what happens when a system built on a shaky definitional foundation also lacks the basic supervisory review to catch an extra six zeros.

    The Bureau’s official response, dated September 15, 2026, is a tidy stack of bureaucratic present-perfect tense (standards issued, certifications scheduled to begin October 1, training “planned” for fiscal year 2027, a Capital Investment Plan that reaches all the way out to 2031). The BOP requested that the OIG close two of the five recommendations outright. The OIG agreed to close exactly one, the definitional standard, and left the other four open until it sees actual documentation that the certifications happened, the training occurred, and the refrigeration units got replaced rather than merely scheduled. Bureaucracy Times has covered this instinct before; an agency describes its own promises as accomplishments and asks the watchdog to take its word for it. The OIG, to its credit, did not.

    Where the Money Actually Goes While Nobody Is Counting

    This audit lands squarely inside the piece we have been building on the BOP’s $5 billion supplemental funding under P.L. 119-21, and it sharpens the stakes considerably. Three billion of that money is meant for hiring and retaining Food Services and Facilities staff, the same staff who, according to this report, often lack basic training in the very system meant to track the equipment they are hired to maintain. Auditors found Facilities personnel at FCC Butner who did not know how to enter routine work orders into TMS and instead filled out paper forms for a supervisor to input on their behalf. That is not a training gap so much as a training absence, extending back years, by the Bureau’s own account, due to lack of budget.

    Congress is being asked to trust that an agency will spend its new money wisely, on the strength of a system that once believed a griddle cost more to maintain than most small businesses are worth. The BOP says the fix is coming. The OIG’s response, essentially, is: show us the paperwork first, and this time make sure someone checks the math.

    Fediverse Reactions

  • Amtrak Was Born a Fiction, and Fifty Years Later Nobody Has Fixed It

    A For-Profit Company Nobody Believed Would Turn a Profit

    Every government corporation in this series has had to pretend, at some level, to be a real business. Amtrak is the only one where the pretending was written into the planning documents. When Congress drafted the Rail Passenger Service Act of 1970, internal analysis explicitly described the new National Railroad Passenger Corporation as a corporation organized for profit.

    One of the planners, John McClellan, later admitted what everyone involved already knew: most of the planning team was dubious of the “for profit” claim, but neither the White House nor the more conservative members of Congress were willing to sign off on an entity that was set up from the start to be a perpetual ward of the state, according to an Eno Center for Transportation history of the Act’s drafting.

    Amtrak began operating on May 1, 1971, taking over the common carrier obligations of private railroads that had been losing money on passenger service for decades, in exchange for the right to use those railroads’ tracks at incremental cost, according to the Federal Railroad Administration. It has never once turned a profit in fifty five years of operation, and the fight over whose fault that is has been running, essentially unchanged, for almost the entire time.

    The Same Question, Asked Since 1979

    GAO has been auditing Amtrak’s finances since the agency’s early years, and the title of its 1979 review says everything: Should AMTRAK’s Highly Unprofitable Routes Be Discontinued? By 1995, GAO was still laying out the same three options it always lays out: eliminate the subsidy and privatize the railroad, which GAO itself conceded was unlikely to work because no Amtrak route earns enough revenue to cover its own costs and few private firms would take on that risk; realign the national route network around the most cost effective corridors; or keep funding the whole thing and hope efficiency gains eventually close the gap.

    They never have. By 1998, GAO found Amtrak had consumed roughly 21 billion dollars in federal operating and capital funding since 1971 while reducing its annual net loss only marginally, from 834 million dollars in fiscal year 1994 to 762 million in fiscal year 1997, with losses already projected to climb back toward 845 million the following year.

    By 2020, cumulative federal subsidies to Amtrak had passed 50 billion dollars, according to a policy analysis from the Cato Institute — a figure that predates the 2021 infrastructure law’s further 66 billion dollars in rail funding, part of it going directly to Amtrak and part to state rail grants. No independent source has published a clean post-2021 cumulative total, in part because the infrastructure law’s Amtrak-specific allocations are disbursed over time and tracked separately from annual operating subsidies.

    The 2025 to 2026 reauthorization debate is, structurally, the same conversation GAO was already having in 1979: current testimony from the Department of Transportation cites GAO findings that long-distance routes account for 15 percent of Amtrak’s riders but 80 percent of its financial losses, and calls once again for ending subsidies to underperforming long-distance service and separating Northeast Corridor infrastructure into its own governance structure.

    The Ledger That Blames the Trains Nobody in Washington Rides

    Here is where it gets genuinely strange, because Amtrak’s own numbers do not agree with themselves. Amtrak’s internal accounting system, called Amtrak Performance Tracking, assigned the Northeast Corridor a 352 million dollar above the rail profit in fiscal year 2025, while attributing a 622 million dollar loss to the long-distance inter-regional routes that serve the rest of the country, the same routes GAO and DOT keep flagging as the problem.

    But a Trains Magazine analysis argues this allocation runs essentially backwards, and it quotes two former Amtrak chiefs to make the point. Former CEO Wick Moorman told Congress in writing that eliminating the entire long-distance network would actually increase Amtrak’s total loss and subsidy requirement by 435 million dollars a year, meaning those routes are net contributors once the Northeast Corridor’s fixed costs are honestly allocated rather than parked somewhere else in the ledger.

    Former Amtrak President David Gunn reportedly went further, warning that if Congress ever gave Amtrak less than its full subsidy request, the railroad would have to close the Northeast Corridor before it touched the long-distance routes.

    Whether or not you buy the full argument, the fact that Amtrak’s own former leadership and Amtrak’s own official accounting cannot agree on which half of the network is actually losing money, after fifty five years and one accounting system built for exactly this purpose, tells you something about how useful that accounting system actually is for anyone trying to make an honest funding decision.

    It also tells you who benefits from the current framing. That is, the Northeast Corridor serves the members of Congress who appropriate Amtrak’s budget every year, and an accounting method that quietly shields it from scrutiny while pointing the finger at rural long-distance service is, at minimum, a very convenient place for the numbers to land.

    Fifty Billion Dollars and Counting

    Even the number Amtrak takes to Congress each year is a chosen number rather than a neutral one. Amtrak requests funding based on what it calls its adjusted operating loss, 598.4 million dollars in fiscal year 2025, a figure that specifically excludes roughly a billion dollars a year in depreciation, 364 million dollars in capital project planning costs, and hundreds of millions more in expenses that sit on the operating side of the ledger under ordinary accounting rules but that Amtrak has decided are not relevant to the number it wants Congress to see.

    None of this makes Amtrak unique in wanting to look better on paper than it is. What makes it worth closing this series’ current stretch on is how openly the fiction was built in from day one. TVA and the RFC grew into unaccountable financial structures over time, as the incentives shifted. The FDIC‘s clean design cracked under a genuine stress test nobody had planned for.

    Amtrak was chartered as a for-profit corporation by people who wrote down, at the time, that they did not believe it would be profitable, because admitting the truth would have meant admitting they were building a permanent federal dependent. Fifty five years, more than 50 billion dollars in direct subsidies before the 2021 infrastructure law, and one internal accounting system whose own former executives cannot agree with it later, that is exactly what they built.

    Next in this series: the Pension Benefit Guaranty Corporation, the quiet federal backstop between millions of workers and the pension promises their employers could not keep.

    Fediverse Reactions

  • Fed Minutes Reveal AI Boom Is Becoming a Monetary-Policy Problem

    The Federal Reserve once viewed artificial intelligence primarily as a potential productivity engine. Its latest meeting minutes show policymakers confronting another side of the boom: AI investment is contributing to inflationary pressures, enormous borrowing demands and higher costs across the economy.

    The artificial-intelligence boom has grown large enough that the Federal Reserve is no longer merely wondering how much more productive it might make the American economy.

    Officials are now considering how much inflation it might create along the way.

    Minutes from the Federal Open Market Committee’s September 15–16 meeting show artificial-intelligence investment threaded through the Fed’s assessment of inflation, business investment, wages, corporate borrowing and even Treasury yields. Most significantly, some policymakers explicitly connected AI-related demand with the case for higher interest rates.

    The committee unanimously raised the federal-funds target range by a quarter percentage point to 3.75 to 4 percent at the September meeting. All participants supported the increase, citing inflation that remained elevated, a labor market near full employment and an economy continuing to expand at a solid pace.

    The rate increase itself was already known.

    The minutes reveal considerably more about what was happening underneath it.

    An AI boom with an inflation bill

    Fed staff estimated that headline personal consumption expenditures inflation reached 3.8 percent in August, while core inflation remained at 3.4 percent.

    The causes were not confined to the familiar suspects.

    Staff attributed the increase largely to previous tariff increases, higher energy and input costs resulting from geopolitical developments and an increase in prices for technology-related consumer goods associated with the AI buildout.

    FOMC participants went further.

    They said surging AI-related investment was contributing to inflationary pressures and noted that effects from the AI buildout appeared to be increasing in core goods even as the effects of earlier tariff increases faded.

    Businesses were also encountering higher transportation and input costs. Some policymakers warned that AI investment could become large enough for aggregate demand to outpace aggregate supply, creating additional upward pressure on prices.

    That concern matters because the Fed is confronting an economy that has stubbornly refused to provide the traditional justification for easier monetary policy.

    Unemployment fell to 4.1 percent in July and August. Payroll gains accelerated in August. Consumer spending strengthened. Business investment remained robust.

    And once again, AI was part of the explanation.

    Fed staff said the AI buildout was supporting substantial increases in business investment and fueling imports of high-tech capital equipment.

    In other words, the same investment boom promising greater productive capacity tomorrow is adding demand to an already strong economy today.

    From productivity story to policy consideration

    That tension has become progressively more visible in the Fed’s public record.

    Earlier this year, AI frequently appeared in Fed discussions as a source of potential productivity growth and economic disruption. By June, policymakers were discussing strong demand for AI infrastructure as a source of upward pressure on technology and electricity prices and considering whether unusually strong investment could contribute to persistent inflation.

    By July, the Fed’s Monetary Policy Report was documenting rising prices for semiconductors and other data-center components alongside growing demand for industrial metals and AI-related manufactured goods.

    The September minutes move the discussion another step forward.

    Some participants worried that the longer inflation remains above the Fed’s 2 percent target, the greater the possibility that higher inflation becomes embedded in expectations and wage and price decisions. Others were concerned that AI-related demand could produce broader price pressures rather than remaining confined to technology markets.

    Then comes perhaps the most consequential passage in the minutes.

    A couple of participants said a higher policy rate would help prevent sector-specific price increases caused by energy disruptions and AI-related demand from spreading into more persistent inflation.

    That does not mean AI caused the September rate hike. The minutes provide no basis for such a conclusion. Persistent underlying inflation, energy prices, geopolitical instability, tariffs and continued economic strength all figured into the committee’s deliberations.

    But the minutes establish something more subtle and potentially more important: Federal Reserve policymakers are now explicitly considering AI-generated demand when determining how restrictive monetary policy needs to be.

    AI is competing for capital, too

    The effects are not confined to consumer prices.

    The Fed’s market discussion points toward another consequence of the enormous sums being spent on AI infrastructure: competition for capital.

    Treasury yields rose about 35 basis points across the two- to 10-year portion of the yield curve during the period between meetings. Higher expectations for monetary policy and strong economic data accounted for part of that increase.

    But market commentary identified another factor: heavy private-debt issuance being used to finance AI infrastructure.

    According to the minutes, that borrowing was contributing to competition for capital and was among the factors associated with higher Treasury term premiums and yields.

    The scale is significant enough that the Fed separately discussed the corporate debt issued by AI “hyperscalers.” Yield spreads on that debt remained wide because of the large volume being issued and the relatively long duration of the securities.

    At the same time, companies benefiting directly from AI infrastructure investment were outperforming the broader stock market.

    The result is an unusual economic feedback loop.

    AI companies and their partners spend heavily on data centers, chips, electricity infrastructure and other equipment. That investment supports economic growth and corporate earnings, but also increases demand for goods, skilled labor, energy and capital. Those pressures can raise prices and borrowing costs. Persistent inflation then gives the Federal Reserve reason to keep interest rates higher.

    Higher rates, in turn, raise borrowing costs throughout the rest of the economy.

    Housing provides a particularly stark contrast. While financing remained broadly available to larger businesses, the Fed described conditions as restrictive for mortgage borrowers and small businesses. Home-purchase borrowing remained depressed.

    The AI investment boom, in other words, is occurring in an economy where access to capital is far from uniform.

    Pay now, productivity later

    None of this means the Federal Reserve believes AI will ultimately be bad for the economy.

    Quite the opposite.

    Participants generally expect AI investment to increase productivity and potential economic output in coming years. But they also acknowledged considerable uncertainty about the size and timing of those benefits. Some additionally warned about cybersecurity and other risks associated with rapid AI adoption that could offset productivity improvements in certain circumstances.

    That creates the central paradox buried in the September minutes.

    The American economy may eventually receive a substantial productivity dividend from artificial intelligence. But building the infrastructure necessary to produce that dividend requires enormous amounts of capital, equipment, energy and skilled labor today.

    And the Fed has to conduct monetary policy in today’s economy, not the economy AI proponents expect several years from now.

    Fed staff now expects inflation to decline gradually and finally reach the central bank’s 2 percent objective in 2029. Its inflation forecast for 2026 through 2028 was revised upward from July, while its outlook for economic activity and the labor market strengthened. Staff also concluded that inflation risks remained tilted to the upside.

    Most FOMC participants consequently believed another rate increase would probably be appropriate before the end of the year, although they emphasized that future decisions would depend on incoming economic data.

    The committee’s September statement summarized the situation much more simply: economic activity remained solid, capital investment was robust and inflation remained elevated.

    The minutes released weeks later supply a detail missing from that abbreviated public account.

    Artificial intelligence is no longer sitting somewhere off in the Fed’s economic forecast as a hypothetical future productivity revolution.

    It is already showing up in prices, investment, corporate borrowing and financial markets.

    And now it is showing up in the interest-rate discussion.

    Sources

    Fediverse Reactions

  • Twenty Two Years, Three Audits, One Excuse.

    The Social Security Administration runs a program that pays benefits to the teenage children of dead, disabled, and retired workers, provided those teenagers are still in high school. The mechanism that decides whether the checks keep coming is a single form, filled out by the student, certified by a school official, and reviewed by an SSA employee who is trusting both of them to be right. It is not a complicated system. It has also failed the same audit three separate times since 2004, for the same three reasons, with SSA agreeing to fix it on all three occasions.

    A Teenager’s Signature Is Still the Control

    Under the Social Security Act, a minor child of an insured worker loses benefits at 18 unless they qualify as a full time student, in which case payments continue until they graduate or turn 19, whichever comes first (42 U.S.C. §§ 402, 423). The instrument that governs this is Form SSA-1372, Advance Notice of Termination of Child’s Benefits. The student states their school, their attendance dates, and their expected graduation date; a school official signs off; SSA takes it from there.

    There is no automated feed from any school system telling SSA when a kid drops out, transfers, or graduates a semester early. There is no real time check against enrollment records. There is a piece of paper, filled out months in advance, describing a future the student is predicting rather than reporting. The Office of the Inspector General’s newest audit (report 022515, released September 2026) exists because that paper, and the employees who process it, keep getting it wrong the same three ways, and this is the third time OIG has documented it happening.

    2004: The Original Diagnosis

    In an August 2004 review titled School Attendance by Student Beneficiaries over Age 18 (A-09-04-14013), OIG sampled 300 students out of a population of 254,121 and found $202,917 in incorrect payments and $56,065 in unsupported payments among them. Projected to the full population, that came to about $70 million in incorrect payments to 32,839 students and $39.5 million in unsupported payments to 10,312 students. The stated causes were that students and schools did not promptly report attendance changes, and that SSA did not retain the documentation needed to prove eligibility either way. SSA’s response was to promise additional employee training and a systems modification to stop paying benefits past the maximum age of entitlement, a related failure mode significant enough that OIG had already given it a separate audit of its own the same year.

    2014: The Same Chart, Bigger Numbers

    A decade later, OIG returned with Payments to Student Beneficiaries (A-09-13-13059, December 2014), sampling 275 students out of 1,539,078 who had received student benefits between 2007 and 2013. The estimate had grown to $225.3 million in overpayments to about 106,000 beneficiaries and $968 million in unsupported payments to roughly 246,000. The trade press covered it as a program in disarray, and the underlying diagnosis had not moved an inch from 2004. Students and schools were still not reporting attendance changes. SSA employees were still failing to retain evidence, and the agency’s own systems still weren’t recording graduation dates on the Master Beneficiary Record where an employee could see them. SSA’s corrective action, according to the background section of this year’s report, was to notify schools and organizations about the importance of timely reporting and remind its own staff to keep the paperwork. In practice, that was a memo.

    2026: Reading the Same Homework Back to Itself

    The current audit sampled 100 of 521,393 beneficiaries entitled to student benefits from January 2022 through December 2024. SSA correctly paid 71 of them. It got 17 wrong and could not document eligibility at all for 12 others, a documentation gap that on its own accounts for more projected dollars than the confirmed errors sitting next to it, an estimated $342 million in unsupported payments to 63,000 beneficiaries against $211 million in confirmed improper payments to 89,000.

    Split the 17 error cases and the causes fall into the same two buckets OIG has now written up three times. Nine cases trace back to students or schools handing SSA outdated or simply wrong attendance information, the kind of error where a beneficiary reports an expected June graduation and then quietly leaves school the previous January, collecting five months of benefits nobody was still entitled to. The other eight trace back to SSA itself. In one, an employee typed a graduation date into the Master Beneficiary Record a full year earlier than what the student had actually reported, cutting off seven months of benefits the student should have received. In another, a beneficiary attending a full year school got their benefits terminated a month early because an employee applied the rules for a quarterly academic calendar instead, a distinction the agency’s own manual spells out clearly enough (SSA, POMS, RS 00205.325, C). In a third, SSA received written notice from a school that a student had graduated a month ahead of schedule and simply never acted on it.

    The report closes with three recommendations. Fix the eight remaining attendance error cases. Fix the six remaining duration miscalculations. Identify why any of this keeps happening at all. SSA’s Chief Risk Officer replied that the agency agrees. That sentence, or one close enough to it, is now sitting in the file from 2004, from 2014, and from 2026.

    What has never been sitting in that file, across three audits and twenty two years, is a system that checks a school’s actual enrollment against what a form predicted six months earlier, or a Master Beneficiary Record built so a graduation date gets flagged rather than typed in by hand and trusted. The dollar figures keep climbing because the beneficiary population keeps climbing, not because anyone at SSA found a new and interesting way to get this wrong. The agency has spent over two decades photographing the same crack in the same wall, filing the photograph, and calling the filing a corrective action plan.

    Fediverse Reactions

  • Forty Six Percent of Alaska’s Village Water Systems Have Nobody Watching Them

    Somewhere in Alaska in 2025, tribal members turned on a tap and got water contaminated with E. coli, a bacterium capable of causing gastrointestinal illness serious enough, in vulnerable people, to become life threatening. The cause was not a burst pipe or a natural disaster.

    According to a new Government Accountability Office report on tribal water infrastructure, it happened because the utility did not have enough qualified personnel on staff to make sure its chlorination equipment was actually working (Government Accountability Office, GAO-26-107178).

    This was not an isolated incident. It happened twice in 2025 alone, and it fits inside a much larger, quietly alarming statistic buried in the same report, that nearly half of Alaska Native village drinking water systems had no primary certified operator on staff during that same fiscal year.

    The Job Nobody Explains Until Something Breaks

    Water operators are, in the most literal sense, the reason a water system does what it is supposed to do after the ribbon-cutting ends. GAO’s report spells out exactly what that job entails, running the equipment, controlling the treatment processes, and monitoring the plants that make water safe to drink, while wastewater operators do the equivalent work removing pollutants before water is released back into the environment.

    These are not entry level positions filled on a whim; operators need specialized training and must earn and maintain certifications tied to the specific complexity of the system they run, certifications that exist precisely because getting this work wrong has direct public health consequences.

    A water system without a certified operator is not simply understaffed in the abstract sense that any workplace might be short a hand. It is a facility that treats and delivers drinking water without anyone certified to confirm it is doing so correctly.

    That is the exact condition GAO found across a huge share of Alaska Native villages. Citing State of Alaska data, GAO reports that about 46 percent of Alaska Native village drinking water systems, 73 out of 157, did not have a primary water operator certified to the level required for that system on staff in fiscal year 2025.

    That figure applies specifically to systems that were required to have a certified primary operator in the first place, meaning it is not padded by facilities too small to need one. Nearly half of the systems that were supposed to have a qualified person running them did not.

    Why Nobody Stays

    GAO and the State of Alaska both point to the same root cause, and it is not a mystery requiring further study. Rural utilities in Alaska experience high operator turnover, and IHS senior officials attributed that turnover directly to a lack of tribal funding for water operator salaries.

    This is not a story about a shortage of qualified people willing to do specialized technical work in remote communities; it is a story about tribal water systems that, once built, frequently lack the money to pay the specialized staff needed to keep them running, so operators leave for jobs that pay better, and the system cycles through vacancy after vacancy. One utility, unable to afford the energy costs required to run its wastewater aeration system, simply stopped aerating its wastewater, a direct, documented consequence of an operating budget too thin to cover routine costs.

    This turnover and funding problem sits at the center of one of the report’s core findings, that IHS generally does not fund routine operations and maintenance for the water systems it helps build.

    IHS pays to construct the plant. It largely does not pay to keep qualified people running it afterward. The system gets built, the ribbon gets cut, and then the ongoing cost of the one thing that actually keeps water safe, namely a trained person watching the equipment, becomes the tribe’s problem alone, in communities that GAO’s own broader report describes as chronically underfunded relative to the tax base and financing options available to non tribal communities.

    What an Empty Operator Position Actually Costs

    GAO does not let this stay abstract. In 2025, two tribes experienced E. coli contamination in their water systems specifically because they lacked adequate operations and maintenance capacity, including not having enough qualified personnel on hand to ensure chlorination equipment was functioning properly.

    E. coli contamination in a drinking water system is not a paperwork violation; it is a direct threat of gastrointestinal illness and, in vulnerable populations, more serious health outcomes. In a separate 2024 incident, a different tribe lacked enough wastewater utility operators to perform routine maintenance on a lift station, and when the pump inevitably failed, the result was prolonged sewage overflow.

    These are not edge cases dredged up to make a point. They are what GAO explicitly identifies as the predictable consequence of a system where nearly half of a state’s tribal water utilities cannot keep a certified operator on staff, wired directly to a funding structure that pays for construction but not for the ongoing salaries needed to run what gets built.

    The connection between the abstract 46 percent figure and an actual gastrointestinal illness outbreak is not speculative; it is the same causal chain GAO lays out explicitly, funding gap, operator turnover, unmonitored equipment, contamination event.

    A Pilot Program, Finally Proposed

    GAO’s response here goes further than most of its recommendations elsewhere in the report, because this is the one place where GAO does not simply ask an agency to fix its own paperwork. GAO suggests Congress consider authorizing and directing IHS, working with tribes and in collaboration with EPA, to establish a pilot program that would actually help fund tribes’ routine operations and maintenance costs, specifically things like water operator salaries and the energy costs needed to run treatment equipment.

    It is worth noting what this recommendation implicitly concedes, that the current system, in which the federal government funds construction but not the staffing needed to safely operate what it built, is not working, and IHS itself has acknowledged its statutory authority does not currently extend to covering routine operations and maintenance costs at all.

    A pilot program is a modest first step relative to the scale of the problem GAO documents. Forty six percent is not a rounding error or a temporary staffing hiccup; it is essentially a coin flip on whether any given Alaska Native village’s water system currently has a certified professional watching over it. Every one of those vacant positions is a system running, or not running, on borrowed time, waiting for the next equipment failure nobody catches until people start getting sick.

    Source: U.S. Government Accountability Office, Tribal Water Infrastructure: Federal Agencies Should Better Track and Coordinate Their Assistance to Alaska Native Villages (GAO-26-107178, September 2026). The staffing gap is one of three structural failures in the same report. Tribes also contend with a forty-year statutory prohibition that bars IHS from connecting water to homes built with HUD housing funds, leaving communities to choose between homes without water or fewer homes. And when multiple agencies try to fund projects together, the absence of a standardized USDA-IHS interagency agreement adds nearly $100,000 in overhead per project — overhead that produces no water, no pipe, and no certified operator.

    Fediverse Reactions

  • USPS Said Its Lost-Mail Center Returned 39% of Packages. The Inspector General Found Less Than 1%

    The U.S. Postal Service thought its national lost-package operation was returning roughly four out of every 10 eligible packages to customers.

    Its inspector general took another look at the math. Measured against all packages arriving at the Mail Recovery Center, the return rate was less than 1 percent.

    That is one of the more striking findings in a new USPS Office of Inspector General audit of the Mail Recovery Center, the Postal Service’s Atlanta-based destination for “dead mail” — packages that cannot be delivered to their intended recipient or returned to the sender.

    The audit found that USPS was dramatically underestimating how many packages were entering the operation, using a return-rate calculation that excluded the overwhelming majority of incoming packages and relying on a spreadsheet containing a formula error that had gone undetected since 2016.

    All told, the OIG estimates roughly 36 million packages arrived at the Mail Recovery Center between October 2023 and February 2026. USPS operational figures showed only about 9 million.

    The 27 million-package math problem

    The discrepancy begins with how USPS estimated incoming package volume.

    Because every package entering the Mail Recovery Center was not individually scanned, the Postal Service estimated package counts using the weight of incoming containers. The calculation assumed an average package weighed 3.56 pounds.

    OIG examined actual scan data and found the average was closer to 0.89 pounds.

    That difference is not exactly rounding error. A 200-pound container that would be estimated to contain roughly 56 packages using the Postal Service’s assumption could contain about 225 using the OIG’s observed average.

    Applied across the operation, OIG concluded the Mail Recovery Center received approximately 36 million packages during the audit period — about four times the 9 million reflected in the center’s reported volume.

    Then came the return rate.

    Mail Recovery Center management reported returning about 39 percent of packages to customers. But the denominator used for that calculation included only 607,233 packages the center considered eligible for return, rather than the roughly 36 million packages OIG estimated actually entered the facility.

    That excluded about 98 percent of incoming packages from the calculation.

    Measured against all incoming packages, OIG calculated the customer return rate at less than 1 percent.

    The distinction was not merely academic. According to OIG, Postal Service executives cited return rates in the 30-to-40-percent range believing the figures represented all incoming packages. They did not.

    Most packages entering the center were ultimately auctioned, recycled, donated or discarded.

    And then there was the spreadsheet

    OIG also discovered an error in the Excel spreadsheet used to calculate Mail Recovery Center performance metrics.

    The formula double-counted certain successfully returned packages, further inflating the reported performance figure. According to the audit, the error dated to 2016 and remained undetected until inspectors examined the calculations.

    USPS management continues to disagree with OIG over how the return rate should be presented. The Postal Service argues the existing metric appropriately measures packages that are actually eligible to be returned. OIG argues an overall return rate is also necessary because senior management had interpreted the narrower measure as describing the entire stream of packages entering the facility.

    That recommendation remains unresolved.

    A free disposal service for commercial shippers

    The audit identified another expensive problem hiding inside the mountain of undeliverable packages.

    At high-volume facilities reviewed by OIG, roughly 75 percent of dead-mail packages came from commercial shippers that repeatedly used the Mail Recovery Center by refusing returned packages carrying postage due or otherwise avoiding taking the merchandise back.

    In effect, OIG said, the Mail Recovery Center was providing some commercial shippers with a free disposal service.

    One fulfillment center outside Los Angeles reportedly refused approximately 250 returned packages every week for three to four years rather than pay the return postage.

    OIG estimated USPS lost $109 million in return-postage revenue between October 2023 and February 2026 and calculated that a refusal fee could have generated another $54 million during the period. USPS disputed those monetary calculations, saying it could not replicate OIG’s estimates and arguing that revenue generated from reselling abandoned merchandise should be considered against processing costs. OIG stood by its methodology.

    Meanwhile, customers were waiting

    The operational problems were visible outside the accounting system.

    Missing packages generated approximately 5 million customer complaints in fiscal 2025, more than any other USPS service or product, according to the audit.

    OIG found that only 7.6 percent of packages scanned when they arrived at the Mail Recovery Center had previously received the required scan showing they were being sent there. Without that scan, customers — and sometimes postal employees — could not readily determine where a package had gone.

    Inspectors also found packages looping repeatedly through the postal network because employees failed to mark them properly. One package reached the Mail Recovery Center five separate times over nearly four months.

    As of May, the center also had a four-month backlog of damaged packages awaiting processing. One package mailed from Cleveland on Dec. 19, 2025 — apparently intended as a Christmas shipment — finally reached its customer May 11, 143 days later.

    USPS agreed or partially agreed with eight of OIG’s nine recommendations, which include improving package-volume measurements, strengthening required scanning, addressing repeat commercial refusals and improving performance reporting.

    The audit ultimately describes something larger than a warehouse struggling with lost Christmas presents. The Postal Service had a national operation handling tens of millions of undeliverable packages, but its own measurement system understated the volume by roughly 27 million packages and produced a success rate senior executives misunderstood.

    For an operation devoted to finding things that went missing, the Mail Recovery Center appears to have temporarily misplaced its denominator.

    Fediverse Reactions