The Pilot Was Supposed To End April 4. Instead The SBA Made It Permanent.

Somewhere on a calendar nobody at the Small Business Administration wants to talk about, April 4, 2026 came and went as the scheduled end date of a $300,000 Palantir bootcamp contract. Nothing shut down. On July 14, 2026 the agency announced a new phase instead, converting the pilot into an open-ended partnership and pointing the software at borrower data in California, Ohio, Minnesota, Wisconsin and Maine. A pilot with an expiration date is a test. A pilot that survives its own expiration date and keeps growing is a program, and nobody sent out a press release admitting that the test never actually ended.

Published July 29, 2026 • Filed under: The Pilot That Would Not Die, Five States And Counting

Server racks glowing in a dark data center, standing in for the Palantir infrastructure the SBA quietly kept running past its own contract end date

A contract has a start date and an end date for a reason. The reason is that somebody, somewhere, is supposed to look at the results before deciding whether to keep spending money. The SBA signed its first Palantir agreement in January 2026 for $300,000, structured as a fraud prevention pilot and bootcamp, scheduled to wrap on April 4, 2026. That date is not a rumor or an estimate. It is the actual contractual end point reported at the time the deal was signed. April 4 came. April 4 went. And on July 14, 2026, more than three months past the date the pilot was supposed to close, the SBA announced what it called a new phase of the partnership, not a renewal, not an extension, a new phase, as if the old one had simply graduated into something bigger rather than quietly kept running the whole time nobody was watching the clock.

Administrator Kelly Loeffler put her name on the announcement personally. "No amount of fraud is acceptable, whether it is $10,000 or $10 million," she said, framing the expansion as a matter of principle rather than a contract that had already outlived its own paperwork. The agency's language around the new phase talks about exposing fraudulent actors, supporting criminal enforcement actions, and recovering stolen funds with advanced technology. What it does not talk about is why a three-month bootcamp needed four extra months to decide it liked what it saw, or why the answer to that question arrived dressed up as an accomplishment instead of an overrun.

Five States, One Very Consistent Number

The new phase comes with a scoreboard, and the scoreboard is where this story stops being abstract. Suspensions across five states have now topped 150,000 pandemic-era borrowers, tied to more than $10 billion in suspected fraud. Break it down state by state and the pattern holds with almost mechanical consistency. California carries the heaviest weight, 112,000 borrowers suspended and $8.6 billion in suspected fraud. Ohio follows at 27,000 borrowers and $1.1 billion. Minnesota, the state whose scandal started this entire machine, sits at 6,900 borrowers and $400 million. Maine, the smallest population on the list, still accounts for 1,500 borrowers and $93 million. And then there is Wisconsin, a state that has not appeared in this site's running account of the dragnet until now, at 7,800 borrowers and $375 million in suspected fraud. Five states, five separate suspension waves, one algorithm behind all of them, and the count keeps climbing every time the agency issues a new press release.

A pilot that needs a new phase to justify itself past its own end date was never really being tested. It was being installed, one press release at a time, until installed was simply what it was.

What "New Phase" Actually Means In Contract Language

Strip away the branding and the mechanics are simple. The Fraud Prevention Pilot Program was the label on the $300,000 January contract. The new phase is the label on whatever comes after it, and the agency's own framing describes ongoing efforts to identify, investigate and help prosecute fraud in the Paycheck Protection Program and the COVID-19 Economic Injury Disaster Loan program, language that describes a permanent capability, not a bounded experiment. Nowhere in the public announcement is there a stated end date for phase two. Nowhere is there a dollar figure attached to what the expanded work will cost taxpayers going forward. The first contract had a number and a deadline. The second one has neither, and an agency that moved from a priced, dated pilot to an undated, unpriced partnership in a single press cycle has told you something about its own confidence that the language was never designed to say directly.

The Part The Announcement Does Not Mention

None of the July 14 materials address what happens to the borrowers who were flagged during the pilot phase and are still waiting on an appeal when phase two begins layering more states and more data onto the same system. This site has already tracked what a suspension actually looks like on the ground, from Ohio borrowers stuck in a 45-day appeal window that functions more like a waiting room than a process, to the plain fact that the word doing all the legal work in every one of these announcements is suspected, not proven. Extending the software's reach into Wisconsin does not extend an explanation to the people already caught in California or Maine. It just extends the software.

The Honest Case For Keeping It Running

It would be dishonest to pretend there is no argument for phase two, so here it is, stated plainly instead of buried. An estimated fraud rate as high as 20 percent inside $1.2 trillion in total pandemic aid is not a rounding error, it is a genuine hole in public money, and a tool that can cross-reference banking records, tax filings and loan files faster than a team of human investigators is not inherently wrong to use. If Palantir's software is actually finding real fraud rings instead of just flagging thin paperwork, then letting a working pilot keep working past its original deadline is a defensible use of four extra months, and the counterargument that a deadline is a deadline can sound more like process worship than a real objection to the results. The problem is not that the SBA kept the tool running. The problem is that it did so without telling anyone the old contract had already lapsed, and dressed the overrun up as a milestone instead of naming it as what it was.

The SBA did not extend a deadline. It let one pass, then announced a bigger version of the thing that deadline was supposed to close out, and called the whole sequence a new phase. That is not renewal, it is mission creep with better branding, and the fact that it is happening across five states now instead of one is the clearest evidence yet that this was never going to stop at a bootcamp. If you were flagged, suspended, or referred to Treasury under a pilot program that outlived its own contract, send us the story, and the rest of the agency's surveillance record so far is filed here.

SHARE ON X SHARE ON FACEBOOK SHARE ON LINKEDIN