United States.Immaculate Constellation · alleged UAP-archive uSAP

Starting October 9, 2024, the press broke the story of an alleged unacknowledged Special Access Program regarding unidentified anomalous phenomena. Investigative journalist Michael Shellenberger published a report on his site Public and, above all, gave it institutional reach by handing lawmakers a 12-page whistleblower document, described as the original Pentagon document on Immaculate Constellation. On November 13, 2024, Shellenberger testified at a joint hearing of two House of Representatives subcommittees (of the Committee on Oversight and Government Reform), where he detailed the document's content; the physical copy was displayed by Representative Tim Burchett, and Representative Nancy Mace later published it on her official House page.

Per the document, the program was allegedly created around 2017 and its mission would be to collect, quarantine and analyze high-resolution imagery and sensor data of UAP captured by U.S. military assets worldwide —reconnaissance satellites, air and naval platforms—, managed outside appropriate congressional oversight. The text also describes morphological categories of the recorded objects (spheres or orbs, discs, 'tic-tac' ovals, triangles, boomerang shapes and irregular bodies), a detail the corpus records without endorsing, since there is no verified imagery to back it. The Department of Defense issued a characteristic non-denial response: it stated that no program by that name exists, but declined to explain how the documents describing it came to be written or who originated them.

The file enters the corpus for its documentary and institutional trace —document entered into the congressional record, DNI FOIA reference (DF-2025-00021), official DoD non-denial— and not for confirmation of the phenomenon: the program's existence remains independently unverified.

In 2025 the whistleblower behind the document identified himself publicly: Matthew Brown, a counterintelligence analyst who says he worked for the Department of Defense and the Department of State with TS/SCI clearance. In interviews with the Weaponized podcast (Jeremy Corbell and George Knapp) and with Judicial Watch, Brown said he stumbled onto the files by accident on a shared government server, and spent four years reconstructing their content from memory, notes and fragments —without having kept the original document.

A later critical analysis identified the likely origin of the misunderstanding. According to researcher Jason Colavito and an extensive thread on the skeptic forum Metabunk, the folder Brown interpreted as evidence of a real program was labeled '2018 Schriever Wargame' —the Schriever wargames are biennial U.S. Air Force/Space Force space-conflict simulation exercises with explicitly fictional, future-set scenarios (the 2018 edition was set in 2028). The skeptical reading is that Brown, lacking a military background to recognize the wargame format, mistook a hypothetical planning document written in declarative language for a real intelligence report. Metabunk further notes that his account rests entirely on memory of a file seen years earlier, with no verifiable origin metadata, and that the alleged 'USS Trepang' photographs he cited as corroboration show duplicated cloud patterns, a sign of manipulation.

On August 22, 2026, Brown took part in a Reddit AMA alongside whistleblower Jeff Nuccetelli (see ↗), moderated by journalist Leslie Kean, where he stated —in a variation on his original account— that the program runs out of the White House National Security Council rather than the Pentagon. There is no whistleblower identified under oath before Congress, no released imagery, and no independent documentary corroboration of any part of the account; only the original document handed over by Shellenberger, its partial chain of custody, the official reaction and, now, a well-supported mundane explanation that directly competes with the covert-real-program reading.

It is a case of strong documentary trace on an unconfirmed claim. Unlike a loose testimony or a forum rumor, here there is a physical artifact —a 12-page document— that moved through institutional channels: it was displayed by sitting lawmakers, detailed under testimony at a congressional hearing, referenced in a FOIA response by the DNI itself, and answered by an official DoD non-denial. That ensemble leaves an auditable trace that survives the question of whether the program exists.

The analytically relevant part is precisely the shape of the official response. The DoD did not say 'this is a fraud and here is the proof'; it said no program by that name exists and, at the same time, refused to explain who wrote or where a document came from that describes its procedures in bureaucratic detail. That rhetorical pattern —'it does not exist, but we won't say where the papers came from'— is exactly the motif the corpus tracks in the cover-up→ambiguity cycle: a denial that does not close the question but keeps it open. Honesty, however, requires marking the other side with equal clarity: a document can be authentic in its circulation and still contain false or unverifiable claims; there is no whistleblower under oath, no released imagery, and a text is not evidence of the phenomenon it describes.

That is why it is not Tier S or high probability: the documentary file is real and traceable, but the program's existence —and, far more, the reality of the imagery it supposedly holds— remains independently unverified. Structurally it is comparable to the Wilson-Davis memo of 2002: a contested document of disputed origin, whose value to the corpus is procedural (how the phenomenon's information would be administered and denied) and not evidential (what the phenomenon is).

Matthew Brown's identification as the author and, above all, the Schriever wargame explanation change the case's probabilistic reading: from a contested document with no obvious mundane explanation, to a case with a concrete, well-argued human-error hypothesis —mistaking a fictional planning exercise for real intelligence— though not conclusively closed either (Wikipedia's own citation of Colavito carries a 'failed verification' flag, and Brown has not recanted). It is also a lesson about the corpus's own process: the probability assigned in 2024 reflected what was known then —an institutional document with no competing mundane explanation. The revision with 2025-2026 evidence —not a new fact about the phenomenon, but a better explanation of the document— is exactly the kind of Bayesian update the corpus's MECE model is designed to absorb.