Coupling needs a measure

This continues an earlier piece on the role of trusted connector. That one was about industry. This one is about defence — and the missing component turns out to be the same.

On 4 and 5 August 2026, Valerii Zaluzhnyi — Ukraine's ambassador to the United Kingdom and former Commander-in-Chief of the Armed Forces of Ukraine — set out his view of the coming war twice in public: first at a meeting of ambassadors, then at a diplomatic club in Kyiv.

His diagnosis of the Alliance is blunt: NATO is operating inside a doctrine from the end of the Second World War, while Ukraine is technologically ahead of it. He was careful to add that he is a supporter of NATO, and that in 2019 he was himself the principal advocate of moving Ukraine's armed forces onto NATO standards. The argument is for transformation, not for abandonment.

But the most important part of those remarks was not the assessment of NATO. It was the description of a battlefield nobody is ready for.

A description without a mechanism

Zaluzhnyi describes the coming war as vast digital fields on which robotic systems operate. Then he adds a detail that is easy to read past: those systems may be directed by people, or they may already have an artificial intelligence of their own.

This is not a forecast about the 2030s. It is an “or” that already exists inside every system operating faster than a human can intervene.

He then lists five domains in which capability is required: surface and subsurface maritime, land, air, space, and the digital sphere including cybersecurity. And he draws a conclusion about alliances: the future lies not in treaty articles but in technological coupling between those able to act together.

The same shape as in industry. The role is named. What makes it work is not.

Two languages, one gap

In an August column for the Financial Times, the chief executive of Siemens described Europe as a trusted connector between ecosystems that are drifting apart. The general describes an alliance coupled by technology rather than by articles.

These are different worlds. Civilian industry and defence, different continents of professional experience, different readers. But the structure of the argument is identical: the future lies in connection, not in autarky.

And the missing component is identical too. Connection answers whether systems can interoperate. It does not answer whether they should.

When two people who have not read each other arrive at the same edge and stop at the same point, that is no longer coincidence. It is the shape of the problem itself.

Treaty articles and coupling fail differently

The difference between these two constructions is not reliability. It is the character of failure.

Article 5 fails slowly. Its failure mode is hesitation: consultations, meetings, days of coordination. That is bad, but it happens at human speed, and a human can still intervene.

Technological coupling fails fast. Its failure mode is an action nobody authorised, or an unauthorised refusal to act. And it happens at the speed at which the systems exchange.

A document that takes days to invoke cannot govern a delegation that takes milliseconds. Not because the document is poor, but because these are different orders of time.

An alliance coupled by technology needs a trust mechanism that runs at the speed of the coupling itself. Otherwise there is coupling, but no alliance.

The fifth domain is not fifth

The digital sphere appears fifth in the list, alongside land, sea, air and space. Formally an equal.

In practice it is not parallel to the other four. It is the one through which the other four are commanded.

An uncomfortable consequence follows. A gap in land capability is a gap in land capability. A gap in the digital domain is a gap in all five at once, because command of the rest passes through it.

Which is why the digital domain cannot be measured the way the others are. In the others the question is how much of it we have. In the digital one the question is whether we can prove that it did what it was supposed to do.

An arbiter cannot be a participant

In the earlier piece I argued that neutral measurement is impossible for whoever owns the thing being measured. A platform cannot be a neutral arbiter of trust in the agents that run on it.

Inside an alliance the constraint is harder still. An arbiter cannot be a participant.

If the measure of trust between coalition systems is set by the coalition's largest member, that is not a measure. It is policy with numbers attached. Smaller participants either accept it or stay outside the coupling — and in both cases the alliance rests on asymmetry rather than on trust.

This is an old problem of compliance verification in coalitions, and it has no new solution. It has only the old one: the party doing the measuring must have no stake in the result.

Experience without a seat

Here I have to say something about my own country that I do not enjoy saying.

Zaluzhnyi puts a direct question to the Alliance: is it prepared to fight Russia without Ukraine's experience. The question is rhetorical and the answer is obvious.

But something less comfortable follows from it. Experience is raw material. A standard is a product. And there is no automatic path from one to the other.

A country can hold the world's largest body of operational experience in autonomous systems and still have no voice in how that experience is formalised into specifications. Its experience then becomes the empirical base for standards written elsewhere, under rules it had no part in setting.

This is not a conspiracy. It is simply what happens when some are fighting and others are sitting in committee.

The venues are already opening — among them the ITU focus group on trust and identity for humans and agentic AI, established in July 2026. Being present there costs incomparably less than the consequences of absence.

What follows

Zaluzhnyi's thesis about alliances coupled by technology is correct. But the word “coupled” in it is not a metaphor. It is a technical requirement, and adapters alone will not satisfy it.

What is needed is a measure of trust that belongs to none of the participants, runs at the speed of the coupling, and knows how to refuse when the evidence is insufficient.

Building such a measure is harder than agreeing on an exchange format. It is also the only way an alliance coupled by technology stops being a figure of speech.