REQUIEM

Britain has built a dependence it has never named. The systems that fuse its sensor data and inform its commanders are, in the main, not British. They are built abroad and maintained abroad, and their priorities are set abroad. They can be switched off by decisions in which Britain has no vote.

Cognitive infrastructure is different in kind from hardware. A delivered aircraft keeps working without its maker. AI systems are living software — retrained, updated and evolved by the teams that own them. A nation that relies on a foreign platform does not buy a tool once; it enters a permanent relationship. It cannot see the roadmap. It cannot audit the training data. It cannot set the priorities.

Cognitive sovereignty is the capacity of a state to read its own strategic environment, process its own intelligence without foreign mediation, and reach its own judgements. Its loss is quiet. It accumulates in contracts and habits, and by the time it is visible it is structural.

Britain’s emerging targeting architecture is designed to connect sensors, deciders and effectors on common digital foundations. Two of the three tiers are well understood: Britain knows how to collect, and it knows how to strike. The hard problem sits between them — the decider tier, where fused intelligence becomes decision at machine speed, under contest.

That tier has been named but not yet specified, and it is the tier most likely to default to the most capable system on offer. Today that system is foreign. The architecture is still being set; once a vendor is embedded and workflows form around it, the choice is gone for the life of the programme.

REQUIEM is conceived as a six-layer system on a sovereign foundation: open-source software on UK-controlled infrastructure, with Crown-owned code and no foreign component able to disable it.

Data is ingested from heterogeneous sources and fused into one operating picture. A knowledge graph maps the entities and how they connect. An analytical core adds reasoning, including causal reasoning. A decision-support layer produces options, and above them an Operational Frame. A presentation layer pushes outputs to analysts, commanders and the tactical edge through standards already in service — the relevant NATO standards and the open Cursor-on-Target format. Interoperation runs through open standards; dependence runs through none.

The design rests on three capabilities the deployed systems lack. Causal reasoning, not mere correlation: modelling why a pattern occurs, not just that it recurs — a system that only matches patterns is reliable until an adversary purposely breaks the pattern. Absence intelligence: the detection of the conspicuous gap — the convoy that does not run, the transmitter that falls silent, the rotation that does not happen. And the Operational Framing Engine, whose main output is not a menu of options but a living frame of the adversary’s situation, intent, pressures and likely reasoning. A commander given only options depends on the system to have asked the right question. A commander who holds the frame can ask his own.

REQUIEM advises; the commander decides. Human authority is built into the architecture. Every output carries its provenance. Uncertainty is shown, not hidden.

REQUIEM is a concept, argued on its merits, its evidence open to inspection, its limits stated without flinching. The honest position on the full system is a five-to-seven-year horizon, and its deepest capabilities are research frontiers.

Software development has begun on a bounded, end-to-end vertical slice: open-source ingestion through to a commander’s Operational Frame, on a synthetic theatre scenario, with the absence-intelligence and framing capabilities working in code. Classified data, deeper modelling and full integration are explicitly out of scope for the proof, deferred to later phases.