The Gulf of Oman Cold Orbs: An AC-130 Gunship, September 2021

On the night of 8 September 2021, a U.S. Special Operations AC-130J gunship was flying a live-fire training sortie over the Gulf of Oman. The crew dropped a water-activated flare — a burning marker on the sea surface, used to give the gunners something to shoot at — and shortly afterwards reported two objects they could not identify.

The objects were described as stationary, roughly four feet in diameter, sitting between zero and twenty feet above the water. Over the course of the sortie the crew reported around twenty-five instances of similar phenomena on the aircraft’s infrared sensor, and gave them a name that has stuck: cold orbs. Their visual estimates of speed ranged from 250 to 1,300 miles per hour. The report describes the objects as manoeuvring aggressively, flying in various formations, and appearing to react to the deployment of the aircraft’s onboard weapons.

It is, on its face, one of the more dramatic modern military encounters in the declassified record. It is also a case where almost everything that would let anyone check it is missing, and the government’s own paperwork says so.

The recorder was not working

The single most important fact in the file is administrative. The aircraft’s digital video recorder was not functioning during the incident, so the sortie produced no native sensor data at all.

What exists instead is six short video clips — between sixteen and twenty-six seconds each — recorded on a cellular phone pointed at the infrared sensor display in the aircraft. The Department of War is explicit about the consequences: this footage does not represent primary sensor data, and the secondary recording method introduces limitations in fidelity, resolution and overall visual quality, with blurring and flickering artefacts inherent to filming a screen. Two of the six clips substantially overlap in content, so the six files show fewer than six distinct views.

This matters more than it might sound. Frame-level analysis of a genuine sensor feed can recover a great deal: calibrated intensity values, sensor mode, field of view, aircraft attitude, timestamps, and the metadata overlay that lets an analyst reconstruct geometry. A phone recording of a monitor preserves almost none of it. The image has been through a lens, a display, a second lens, and a lossy codec, and every one of those steps adds artefacts that look like phenomena. The most striking evidence in the case has been degraded, before analysis began, into a form that cannot really be analysed.

What the numbers can and cannot say

The reported speed range is 250 to 1,300 miles per hour. That is not a measurement with an error bar; it is a spread of more than five to one, produced by eye, at night, over open water, against objects of unknown size at unknown distance.

Judging speed requires range, and judging range requires size. Over a featureless sea at night there is nothing to supply either. A crew told that an object is roughly four feet across will derive one set of speeds; the same visual data, assumed to be a larger object further away, yields entirely different ones. The five-to-one spread is not a defect in the crew’s professionalism — it is what honest estimation looks like when the necessary references are absent, and the fact that they reported a range rather than a figure is a point in their favour.

The four-foot diameter has the same problem. It is not a measurement. It is an impression of angular size converted into a physical size by assuming a distance.

The flare

The sequence is worth stating plainly: the crew dropped a burning marker onto the water, and then began seeing objects hovering zero to twenty feet above the water near it.

An infrared sensor pointed at an intense heat source on a reflective surface is operating in close to the worst conditions available to it. Bright sources bloom, producing halos and secondary blobs in the optical path. Sensor auto-gain adjusts to the brightest object in frame, which changes how everything else renders from moment to moment. Water reflects, and a burning flare on a moving swell produces a shifting field of hot and cold contrast. The objects were characterised as cold — darker than their surroundings on the display — and a cold blob near a very hot source is exactly the signature that thermal artefacts around a bright emitter tend to produce.

That does not establish that the cold orbs were artefacts. It establishes that a flare is present in the sequence and is a strong candidate for at least some of the twenty-five instances, and that ruling it out would require the native sensor data that does not exist.

The same reasoning applies to the most arresting claim in the report — that the objects appeared to react to the aircraft’s weapons. A firing AC-130 produces muzzle flash and heat directly in the sensor’s environment. Anything that changes sensor gain at the moment the gun fires will make everything in frame appear to move or change simultaneously. An apparent reaction to weapons fire is precisely what a sensor artefact would look like, and precisely what a genuine reaction would look like too. With a phone recording of a screen, there is no way to distinguish them.

Parallax deserves a mention as well. The gunship is orbiting. Objects at different ranges appear to move relative to one another as the aircraft moves, and objects that are genuinely stationary can appear to track, hold station, or accelerate depending on the geometry. This is the single most common source of dramatic apparent motion in aircraft sensor video, and it cannot be assessed without knowing the aircraft’s flight path — which the release does not contain.

What the government says about its own report

The written half of this case is an Intelligence Information Report, released alongside the footage as DOW-UAP-D101. The format matters. An IIR is the standard vehicle for disseminating unevaluated information from human sources. It is explicitly not a finished intelligence product; it is raw input for later all-source assessment. Publishing an IIR means publishing what somebody reported, not what the government concluded.

The report says so itself, in a caveat that deserves quoting for how carefully it is worded: all descriptive and estimative language reflects the reporter’s subjective interpretation at the time of the event, and such characterisations should not be interpreted as a conclusive indication of the presence or absence of any intrinsic object features or performance characteristics.

Read that against the phrases it is attached to — cold orbs, manoeuvring aggressively, reacting to weapons deployments — and the structure of the case becomes clear. Those are the crew’s words for what they experienced, preserved accurately, and the government has taken care not to endorse them as descriptions of what was physically present.

Why it is still worth having

Nothing above says the crew saw nothing. An experienced aircrew on a live-fire sortie, logging around twenty-five instances across a night and reporting them formally, is not a trivial data point, and the case remains unresolved by the All-domain Anomaly Resolution Office.

What the case actually demonstrates is how thin the evidentiary margin is even in the strongest-sounding modern encounters. Here there were trained observers, a military-grade infrared sensor, a formal reporting chain, six pieces of video, and an official document — and the recorder failure removed the one thing that would have made any of it checkable. The written report and the video have been released together, which is rare and genuinely useful, and reading them against each other is instructive precisely because they do not resolve the question.

That is the honest shape of it. The best modern cases in the archive are rarely undone by a single obvious explanation. They are undone by missing data, unrecoverable geometry, and estimates made under conditions that make estimation unreliable — and by the fact that the most compelling detail in the file is usually the one nobody could measure.

Sources

• Intelligence Information Report, Gulf of Oman — the written report, PURSUE Release 05
• Department of War UFO Disclosure Portal — the six released video clips
• AARO (All-domain Anomaly Resolution Office) — the office that designates the case unresolved