From Battlefield to Courtroom: Defence Procurement's Real Evidence Problem

A guided munition leaves the rail. A drone feed records the strike. Eighteen months later, in a courtroom or a board of inquiry, someone asks a simple question. Is this the same footage that informed the decision, and can you prove no one touched it in between? In most procurement programmes the honest answer is no. The platform was bought to deliver effect on the battlefield. Nobody bought the chain of custody that the effect would later be judged against.

This is the gap that defence ministries keep paying for twice. First in the acquisition, where the sensor, the network and the weapon are specified to exacting standards. Then again in litigation and inquiry, where the data those systems produced turns out to be inadmissible, unsigned, or impossible to distinguish from a plausible edit. The evidence problem is not a legal afterthought. It is a procurement requirement that almost no requirement document contains.

Why the data does not survive the journey

Battlefield data is born in the worst possible conditions for evidence. It is generated at the edge, on contested networks, by systems from a dozen vendors that were never designed to agree on a common record. It is copied, compressed, relayed through ground stations, and ingested into intelligence pipelines that prioritise speed over custody. Every hop is a place where the link between the original observation and the stored file quietly breaks.

By the time the data reaches a place where it could be examined, it has usually lost the three things a tribunal cares about most. It has lost a verifiable point of origin, so no one can swear when and where it was captured. It has lost integrity, so no one can show it was not altered. And it has lost continuity, so the line from observation to exhibit is full of unexplained gaps. Storage was never the issue. A petabyte of unprovable footage is worth less than a single signed frame.

The procurement blind spot

Procurement frameworks are very good at specifying capability and very poor at specifying accountability. A requirement will demand a sensor resolution, a data rate, a network latency, an encryption standard for confidentiality. It will rarely demand that each consequential record carry a tamper-evident, independently verifiable proof of who produced it and that it has not changed. Confidentiality and integrity are treated as the same problem, when they are opposites. One hides the data. The other must let an outsider check it years later without trusting the people who hold it.

The result is a structural asymmetry. The state spends to create overwhelming operational advantage, then finds it cannot account for how that advantage was used when an ally, a court, or its own public asks. In an era of contested narratives, where an adversary will publish a doctored clip within the hour, an unprovable archive is not just a legal weakness. It is an information-warfare vulnerability bought at great expense.

What a sovereign record actually requires

Fixing this does not mean trusting a new central authority with the evidence. That simply moves the single point of failure. The requirement is narrower and harder. At the moment a consequential action occurs, the system must seal a record of it, sign that seal so the signature can be checked by anyone, and commit a fingerprint of it somewhere that cannot be quietly rewritten. Custody then becomes a property of the record itself, not of the organisation holding it.

This is the function Mickai is built to perform. Mickai is a Sovereign Intelligence Operating System (SIOS), running its specialised AI brains on the operator's own hardware, fully offline-capable, so the analysis of sensitive material never leaves sovereign control in the first place. Every consequential action it takes is written to the Open Audit Record (OAR) and sealed with a signature under FIPS 204 ML-DSA-65, the published NIST post-quantum signature standard. Mickai did not invent that standard. It adopts it, which is the point. The proof a court relies on should rest on open, scrutinised mathematics, not on a vendor's word.

Permanence is the last piece, and it is where the design earns its sovereignty. Mickai anchors a hash commitment of each record to Bitcoin through Pantheon, its own Bitcoin-anchored Layer 1 (native token PAN, fixed supply of five billion). Only a fingerprint is anchored, never the underlying material and never any movement of Bitcoin itself. Anchoring is not spending. The effect is that the existence and exact content of a record at a given time can be checked against the most heavily defended ledger in the world, while the sensitive payload stays sealed on sovereign hardware. A tampered copy fails the check instantly, and the failure is something an outsider can reproduce.

Evidence as a procurement line item

The shift this asks of procurement is small to write and large in consequence. Requirement documents should treat verifiable provenance as a named deliverable, alongside resolution and latency, for every system that produces decision-relevant data. The test is concrete. Take any output, hand it to an adversarial party, and ask them to confirm its origin and integrity using only public methods and no access to the people who hold it. A system that cannot pass that test has not delivered evidence. It has delivered a story.

Mickai's claim to take this seriously is itself on the public record. The architecture is described across 101 filed UK patent applications, around 2,234 claims, owned by Mickai LTD, with Micky Irons (Mickarle Wagstaff-Irons) as the named inventor. The portfolio is held privately by its founder, not licensed out to a buyer's competitors. For a defence customer that distinction matters, because the perimeter that proves your evidence should not also belong to the next bidder.

Defence has spent a decade learning that owning the machines is not the same as owning the outcome. The next lesson is that owning the data is not the same as being able to prove it. The journey from battlefield to courtroom is the one stretch of the kill chain that procurement has consistently refused to fund, and it is the stretch on which reputations, alliances and verdicts now turn. Buy the sensor by all means. Then buy the seal that lets you stand behind what it saw.