The Anduril Ghost-X is a single-rotor reconnaissance drone that the British Army has used to demonstrate a sensor-to-strike link with a separate loitering munition. What follows is confined to what the verified record supports.
What the record confirms
The Ghost-X is manufactured by Anduril Industries, introduced in 2023 as the successor to the earlier Ghost 4. It is a single-rotor VTOL platform, Group 2 class, with a maximum take-off weight of 25 kg and a maximum payload capacity of 11.3 kg. Maximum range is recorded at 25 km, matching its control link range, with an endurance of 1.4 hours. It is electric battery powered, launched and recovered vertically.
Its control and autonomy level is recorded as semi-autonomous, using Anduril’s Lattice artificial intelligence system, and operated by a two-person crew. Its sensor payload is an EO/IR gimbal, with the Trillium HD45LP cited as an example. It is unarmed: the record notes explicitly that the armed sibling in the same family is the Bolt-M, not the Ghost-X itself.
Known operators are the US Army, under the Replicator programme, and the British Army, which has tested the platform. The verified record states that the British Army’s 3rd Battalion The Rifles operationally employed the Ghost-X for forward reconnaissance during NATO Exercise Northern Star in Finland in May and June 2026, at a reported distance of approximately 70 km from the Russian border, and that this reconnaissance data was used to feed targeting information to the Bolt-M loitering munition.
It is worth being precise about the platform family distinction here. The Ghost-X and the Bolt-M are recorded as separate platforms within the same Anduril product family, the former an unarmed ISR asset and the latter an armed loitering munition. The reported exercise therefore demonstrates an inter-platform sensor-to-strike chain, not a single aircraft performing both reconnaissance and strike functions. This distinction is relevant to any assessment of where human decision points sit in the chain, since it establishes that at least two separate systems, and potentially two separate operating crews, were involved.
What this does and does not establish
The verified record confirms that the Ghost-X is a real, fielded, AI-assisted ISR platform, that it has been operationally employed (not merely demonstrated in a controlled trial) by a named British Army unit during a specific NATO exercise, and that this employment involved passing targeting data to a separate armed platform, the Bolt-M. It confirms the use of Anduril’s Lattice AI system as the platform’s decision-support architecture, with a two-person operating crew, indicating that this is not an unmanned command-and-control chain but one with human operators at the reconnaissance stage.
The identified source does not describe the specific nature of the human decision point between the Ghost-X’s reconnaissance data and the Bolt-M’s subsequent engagement: in particular it does not state whether, or how, a human confirmed each target before the Bolt-M engaged. No claim is made here about the degree of human control at the point of weapons release, because the identified source addresses the reconnaissance platform and the exercise context, not the engagement decision itself. It also does not provide a verified account of outcomes from the exercise beyond the stated sensor-to-strike link having been demonstrated.
Why it is being watched closely
Three factors explain why this reported exercise employment matters beyond the platform’s own specification. First, proximity: a reconnaissance operation reported at approximately 70 km from the Russian border is a materially different signal than a domestic training exercise, placing a live AI-assisted sensor-to-strike demonstration close to a contested frontier. Second, the sensor-to-shooter link itself: pairing an AI-assisted reconnaissance platform with a separate armed loitering munition is a concrete instance of the “kill chain” concept frequently discussed in the abstract in defence AI commentary. Third, allied adoption: both US Army (Replicator) and British Army testing of the same platform family indicates this is not an isolated national trial but part of a wider allied interest in low-cost, AI-assisted reconnaissance.
What would change this assessment
Three categories of additional evidence would materially change what can be said about this exercise with confidence:
- Human-control detail. A primary-source account of the specific decision process between Ghost-X reconnaissance output and Bolt-M engagement, establishing what human confirmation step, if any, occurred before any strike.
- Exercise outcome data. Verified results of the Northern Star demonstration, including whether any live or simulated engagement occurred and with what outcome.
- Independent verification. A second, independent source corroborating or updating both the platform specification and the specific exercise account above, since the identified source for this entry is a single reference.
Until such evidence is identified and verified, this article’s conclusions are limited to the platform specification and the exercise account set out above.
Source: Anduril Ghost-X programme entry, UAS reference database (verified 2026-07-29), citing Army Recognition’s reporting on the Northern Star exercise.
Leave a Reply