Two press cycles ago, the SBA announced that it had referred approximately $22 billion in suspected fraudulent pandemic loans to the Treasury Department for collection. A few cycles after that, a task force chaired out of the Vice President's office attached a more granular number to the same pool, calling out 562,000 loans worth $22.2 billion that the prior administration allegedly carved out from earlier referral batches. ACA International, the trade group that watches Treasury collection, ran the higher-resolution version of the same number. Fox Business ran the political framing. Nobody on either side of that story stopped to ask the question that should have led the entire write-up. Where did the 562,000 figure come from in the first place, and what does it mean that it came from the SBA itself.
This is the part of the SBA story that does not fit on a chyron. The 562,000 referrals are not the product of a third-party audit. They are not the product of a forensic clawback unit funded out of the appropriations bill. They are not the result of a private fraud detection vendor running its tooling against the program data. They are the SBA's own internal flagging of its own approvals, generated years after those approvals went out the door, batched up, and handed to Treasury so that Treasury can try to claw the money back through tax offsets, federal salary garnishments, and other mechanical clearinghouse plumbing. The agency that wrote 562,000 bad checks is the same agency that wrote the memo saying 562,000 of its checks were bad. There is no independent disconfirmation step. There is only the press release.
The Closed Loop, Drawn Out Step by Step
Imagine, for a moment, that you ran a small bank that had loaned out $1 billion in 2020. Imagine that four years later you walked up to a federal collection partner, handed over a list of borrowers, and said, "These borrowers, we now believe, were never eligible. We approved them anyway. Please go collect." The first question any reasonable counterparty asks is, on what basis did you flag them, and the second question is, why did your underwriting not flag them at the time. The answer in the SBA's case is exactly what it has always been. The agency leaned almost entirely on borrower self-attestation at intake, ran almost no live IRS or bank verification at the time the disbursements went out, and now in 2026 is performing the verification step it should have run in 2020, retroactively, on a pile of files already paid out.
This is the structural problem. The agency cannot mark its own homework retroactively and have anyone outside the building treat that mark as final. The 562,000-loan figure could be exactly right, or it could be off by half in either direction, and from outside the agency you cannot tell, because the same internal team that approved the loans is the same internal team writing the cover memo that flags them. There is no second pair of eyes in this loop. There is one pair of eyes, and a press release.
The Number Is Round Because The Method Is Rough
$22.2 billion across 562,000 loans averages out to roughly $39,500 a loan. That is a tidy figure. It is also a figure that is suspiciously close to the average EIDL advance and small-PPP loan size in the early waves of the program, which strongly suggests the SBA did not surgically pick out the actual fraud cases. It strongly suggests the SBA pulled a wide net based on coarse signals, like SSN reuse, address reuse, or formation date relative to the disaster declaration, and shipped the entire net to Treasury under a "suspected" label. That is not the same thing as a fraud finding. That is a triage trigger. Triage triggers, when run against an entire program, will catch real fraud and they will also catch a meaningful number of legitimate businesses whose paperwork happened to look lumpy.
This matters because the moment Treasury starts offsetting tax refunds and federal benefits against a "suspected" label, real human beings whose only crime was being a sole proprietor with a Florida address and a 2019 LLC start losing money to the federal government before any due process happens. Treasury offset is fast. The SBA's appeals queue is not.
The agency that wrote the bad check is the only agency telling you the check was bad. The agency that should have caught it before the wire cleared is the same agency standing in front of the cameras saying it caught it. Nobody outside the building has been allowed to grade the homework.
The Vance Task Force Frames the Same Math as a Coverup
The political layer on top of all of this is the claim, run prominently in late April 2026 by Fox Business and the Vance task force, that the prior administration "protected" 562,000 pandemic loans by holding them back from earlier Treasury referral batches. That framing has a shelf life that is exactly as long as your willingness to assume the SBA's internal flag list is the ground truth in the first place. If the list is correct, then yes, somebody held it. If the list is overinclusive, which the round numbers and the program-wide net strongly suggest is at least partly true, then "holding it back" can also be described as "not yet finished verifying it." Both stories use the same primary source. Both stories are stuck inside the same closed loop.
What nobody in either framing is pushing for, in writing, is the only thing that would actually settle the question. An independent forensic audit of a random sample of the 562,000 loans, performed by an entity that is not the SBA, with the loan-level data made available to that entity in unredacted form, and the audit's hit rate published. Until that exists, the country is being asked to agree on a $22 billion claim against borrowers based on the say-so of the agency that approved every borrower in the first place.
What Treasury Is About to Do To Real People
While the cable segments argue about who held what file when, the operational reality on the borrower side is already in motion. Treasury Offset Program intercepts of federal tax refunds, federal salary garnishments, and Social Security offsets do not require a court finding. They require a referral. A referral exists for these 562,000 loans. That means that, starting now, real households whose 2025 federal tax refund was supposed to be a few thousand dollars are going to find that refund offset in part or whole against an SBA loan balance, and the first they hear about it is going to be a Treasury notice in the mail. Some of those households are sitting on legitimate fraud they committed. Some are sitting on a loan that the SBA has now, four years later, decided was suspicious based on a coarse triage signal that nobody outside the building has examined.
The borrower has a right to dispute. The dispute process runs through the SBA. The SBA is the same agency that flagged the loan, that approved the loan, that wrote the press release, that said the round number out loud. The borrower's appeal is heard by the body that originated the dispute. There is no daylight in this process for the borrower to get an independent read.
The Pattern Is Older Than This Round
For anyone who has been watching the SBA story since 2020, this is the sixth or seventh iteration of the same arc. Step one, the agency disburses money fast and skips controls. Step two, Inspector General reports come out estimating fraud in the hundreds of billions of dollars. Step three, the agency refers a subset of suspected fraud to Treasury. Step four, the press cycle moves on. Step five, a new political faction cites the same referral list as evidence of either incompetence or coverup, depending on the faction. Step six, a fresh announcement re-references the same dollar figure with a slightly different headline. The loans never get re-examined by anybody outside the agency. The hit rate is never published. The dollar figure becomes a political number, not an audit number.
What An Honest Statement Would Look Like
If the SBA wanted to actually rebuild credibility on this referral, the statement would read something like this. "We have flagged 562,000 loans, totaling $22.2 billion, as showing one or more risk indicators consistent with fraud at the time of underwriting. We do not represent that all 562,000 are confirmed fraudulent. The basis for each flag, by category, is below. We have requested that an independent forensic auditor draw a random sample of 1,000 loans from this set, reach a fraud finding or non-finding on each, and publish the hit rate. Until that audit returns, Treasury offset on these loans will be limited to cases where the borrower has had an opportunity to respond to the specific risk indicator that triggered the flag." That is what an actual fraud-control culture would say. What the agency said instead was a number, a press release, and a referral form.
The 562,000 figure is, in the most generous read, a rough internal triage list that captured a lot of real fraud and probably some legitimate borrowers along the way. In the least generous read, it is a number that exists because saying a number out loud is politically easier than admitting that nobody in the agency knows the real fraud rate, that nobody has ever known, that the IG estimates have always been ranges, that the program design made the real number unknowable from day one.
The Reason This Site Exists
The reason LOLSBA keeps writing about the same agency from a different angle every week is that this is what the inside of an unaccountable bureaucracy looks like. It is not a movie villain. It is not a single corrupt official. It is a closed loop. It is one agency reporting on itself, citing itself as the source, generating its own evidence, marking its own homework, and then issuing the press release that the next day's news cycle treats as fact. There is no external check, no independent audit on the actual loan-level fraud rate, no random sample published with a hit rate, no acknowledgment that the program design made post-hoc verification structurally impossible to do well.
You are allowed to think the prior administration held the file too long. You are allowed to think the current task force is grandstanding. You are allowed to think both. What you are not allowed to do, if you take this seriously, is treat 562,000 and $22.2 billion as audited numbers. They are not audited. They are an internal SBA referral list that nobody outside the building has been allowed to grade. Until somebody grades it, the press release is doing all of the work the controls were supposed to do.