The April 19, 2023, briefing deck that Seán Kirkpatrick delivered to the Senate Armed Services Subcommittee on Emerging Threats and Capabilities carries a Department of Defense prepublication clearance stamp dated April 17, 2023. It was cleared for open publication. It is an official document. So is the November 19, 2024, briefing that Dr. Jon Kosloski — Kirkpatrick's successor as AARO director — presented to the same subcommittee. Both are posted on AARO's own website. And on several basic questions about the UAP dataset, they say different things.
The discrepancies are not trivial. They sit at the core of what AARO presents to Congress as its empirical foundation — the shape, frequency, and character of what the U.S. military is seeing in the sky.
The morphology problem
Start with shape. The April 2023 Kirkpatrick deck reported that the most commonly observed UAP morphology was orb, round, or sphere, accounting for 52 percent of reports. By November 2024, Kosloski's briefing to the same body placed that category at 31 percent — a drop of 21 percentage points. The 2023 deck listed lights as accounting for just 1 percent of morphology reports. The 2024 deck placed lights at 46 percent, the single largest category.
The April 2023 Kirkpatrick deck reported that the most commonly observed UAP morphology was orb, round, or sphere, accounting for 52 percent of reports.
Both decks cover overlapping periods. The 2023 presentation analyzed UAP reporting trends from 1996 to 2023. The 2024 presentation extended that window to October 10, 2024. An eighteen-month extension of the dataset could theoretically shift the numbers — but a swing of that magnitude, across the two most dominant categories, is not what you would expect from routine data accumulation. It suggests either a reclassification of existing cases, a change in methodology, or a change in what data is being included. AARO has not publicly explained which.
A third slide set, released through FOIA under the request number 23-F-0922, adds a further wrinkle. That document — a UAP reporting-trends slide attributed to AARO and carrying Kirkpatrick's name — lists sphere, round, or balloon shaped at 42 percent of morphology reports, and ambiguous sensor contact at 23 percent, with not specified accounting for another 22 percent. Those three categories alone consume 87 percent of the dataset. Neither the 52 percent figure from the April 2023 open hearing nor the 31 percent from the 2024 hearing appears on this slide. The 23-F-0922 document appears to cover the same general period, but the numbers do not reconcile with either public briefing.
This matters because morphology data is not decorative. It is how AARO characterizes what sensors are recording. If the office cannot produce consistent figures across three documents that draw on the same underlying case archive, that consistency problem belongs in any honest assessment of AARO's analytical outputs.
The Go Fast discrepancy
The morphology figures are not the only place where AARO's own slides diverge. The November 2024 Kosloski deck includes a dedicated analysis of the Go Fast video, the clip of a Navy aircraft tracking an object off Florida's east coast in January 2015. That slide states the video was released by DoD in 2017. The Department of Defense did not officially release the Go Fast video in 2017 — it released it in 2020. The date error appears on an unclassified slide presented to the Senate Armed Services Committee and cleared for open publication.
The 2024 Go Fast slide also presents a tension within itself. It states that the UAP's altitude is assessed with high confidence as approximately 13,000 feet above sea level, at a calculated velocity of approximately 45 mph. Then, in the same set of bullet points, it acknowledges that high-confidence analysis of the object's exact position was not possible given a lack of precise positional data from the observing Navy aircraft. High confidence in velocity depends on knowing position. If position cannot be established with high confidence, the velocity figure carries that same uncertainty — a caveat the slide does not explicitly draw.
The Puerto Rico analysis and its internal contradictions
The most analytically dense section of the 2024 Kosloski briefing concerns the 2013 Puerto Rico infrared video, recorded by a U.S. Customs and Border Protection aircraft over Rafael Hernandez Airport on April 26, 2013. The video had circulated publicly for years and appeared to show a fast-moving object splitting into two and entering the water. AARO's 2024 analysis concluded that two distinct objects were airborne in proximity to each other throughout the observation period, that neither entered the water, that the appearance of high speed was due to motion parallax, and that the objects traveled at wind speed in a straight line over land.
Those conclusions are specific and, if supported by the underlying analysis, substantive. But the briefing presents them alongside details that create internal friction. One slide states that trajectory reconstruction showed the objects traveled southwest at a consistent speed of approximately 3.6 meters per second while slowly descending to approximately 200 meters. Winds at the time were recorded at 7 knots — also approximately 3.6 meters per second — from the east-northeast, which is consistent with AARO's claim that the objects moved at wind speed. That part holds together.
What is harder to reconcile is the simultaneous invocation of two different explanatory mechanisms: motion parallax accounts for the perceived high speed, and thermal crossover accounts for the objects appearing to fade in and out. Both explanations may be correct. But the briefing does not explain how trajectory reconstruction, which depends on establishing actual object position, was conducted with sufficient precision to rule out water entry, while the Go Fast analysis — a simpler geometry problem, one might argue — could not establish position with high confidence due to missing sensor metadata. The office applies different analytical standards to different cases, and those differences are not explained.
The South Asia case and a genuine analytical split
The April 2023 Kirkpatrick briefing included a case that stands as the clearest example of AARO presenting a live contradiction rather than a resolution. An MQ-9 operating in South Asia in 2023 recorded what appeared to be an object followed by an unusual trail. The slide listed, in adjacent bullet points, two competing explanations: the trail appears to be cavitation, similar to those caused during propulsion; and the visible trail is a camera-software artifact caused by video-compression algorithms overlaying a captured image on a previous frame.
Those are not complementary findings. Cavitation is a physical phenomenon associated with real objects moving through a medium at speed. A video-compression artifact is a digital artifact with no physical correlate. The slide marked the case as pending peer-review of mission-partners' analytic findings — a designation that at least acknowledged the analysis was unfinished. Whether peer review was subsequently completed, and what it concluded, has not been made public.
Chain-of-custody and classified network uploads
In May 2026, the Department of War — acting on a March 6, 2026, request from eight members of the U.S. House of Representatives for access to 51 potentially UAP-related records — began releasing materials through its war.gov/UFO portal. AARO identified a collection of responsive materials held on a classified network. According to the release documentation posted at war.gov, many of those materials lack a substantiated chain-of-custody.
Two of the released videos illustrate why that qualifier matters. The video titled Syrian UAP instant acceleration was assessed by AARO as likely derived from an infrared sensor aboard a U.S. military platform operating within the CENTCOM area of responsibility in 2021. But the release documentation states the video was digitally altered prior to its upload to a classified network, and that a user uploaded it to that network in June 2024 — three years after the purported incident date. The Iran formation video carries a similar note: likely a legitimate CENTCOM infrared recording from 2022, uploaded to the classified network in June 2024, also lacking a confirmed chain-of-custody.
Who uploaded these videos, under what authority, and why they sat outside normal military records management for years are questions the documentation raises without answering.
What the pattern suggests
Taken individually, each discrepancy has a potential innocent explanation. Methodologies change. Data gets reclassified. Analysts disagree. Slides contain errors. But across three versions of what is nominally the same UAP-trends dataset, producing morphology figures that do not reconcile with one another; a video release date that is wrong by three years; an analytical standard that varies between cases without explanation; and a chain-of-custody problem affecting materials that Congress asked to see — the accumulation is what demands scrutiny.
AAROstates its mission is to minimize technical and intelligence surprise by adhering to the highest scientific and intelligence-tradecraft standards. That framing, drawn directly from the April 2023 Kirkpatrick briefing, sets the bar. Whether the office's public outputs meet it is a fair question, and right now, the public record does not provide enough information to answer it.
What would change the picture most directly: the methodology documentation underlying each version of the morphology dataset, and a public accounting of what changed between 2023 and 2024 and why. If AARO has that documentation and has chosen not to release it, that is a decision worth asking about. If it does not have it, that is a different kind of answer.