Why the Most Important Conversations in Defence AI Happen Off the Record

There is a version of every major defence conference that gets published: the panel transcripts, the keynote summaries, the press releases. And then there is what actually happened — the bilateral conversations between a ministry official and a technology vendor, the working group that produced a joint position paper, the introduction between a general and a founder that led to a six-month capability evaluation.

The published version is usually less important.

The Problem with Open Sessions

When a senior official from a European defence ministry speaks on a public panel, they are not telling you what they think. They are telling you what they have been cleared to say publicly, in a format that creates no liability and commits to nothing. This is not criticism — it is how institutions work. Public forums are for signalling and positioning. They are not where decisions get made.

The practical consequence is that most defence conferences produce a great deal of content and very little progress. The conversations that would have moved something forward — that required candour, that depended on trust, that needed to happen between the three specific people who all happened to be in the room — did not happen because the format did not create the conditions for them.

"What Belgium needs — what every European defence ministry needs — is not more talking shops. We have enough of those. We need structured forums where the people who make decisions can have frank conversations with the people who build the technology, without it ending up in a press release the next morning. That is a rare thing. When you find it, you protect it."

Theo Francken — Minister of Defence, Belgium

What Changes the Format

The events that produce real outcomes in the defence AI space share three structural features that most conferences lack.

Pre-arranged bilateral meetings

The most valuable thing a well-run summit can do is force the right two people into a room together before either of them has decided they want to be there. The default behaviour of busy senior officials is to attend the sessions on the published agenda and leave. Bilateral meeting programmes override this default — they create structured time for the conversations that would otherwise only happen if both parties already had a relationship.

Chatham House Rule working sessions

Chatham House Rule — participants may use the information from a session, but may not attribute it to named individuals — is not a privacy mechanism. It is a permission structure. It tells the most senior and cautious participants that they can say what they actually think. In practice, the quality of information exchanged in Chatham House sessions is qualitatively different from what emerges in on-the-record panels. For defence AI, where many of the most important developments are operationally sensitive, this matters enormously.

Structured working groups with defined outputs

The difference between a roundtable and a working group is not the number of people. It is whether anyone leaves with an obligation. Working groups that produce joint position papers, that assign follow-up actions, that create documented shared positions — these produce outcomes that participants can take back to their institutions and act on.

The Curation Problem

None of these structural features work without curation. The bilateral meeting between a ministry official and a technology vendor is only valuable if the vendor has been assessed as having relevant capability and the official has been identified as having relevant authority. The Chatham House session only produces candour if the participants trust that the room is composed of people who deserve to hear what they have to say.

This is the hardest problem in event design and the one most conferences do not solve. It requires saying no — to the majority of registration requests, to sponsors whose objectives are misaligned with the audience, to sessions that generate traffic rather than progress. The AI in Defence Summit's curatorial model — identity verification and editorial review of all sponsorship applications — is a direct response to this problem.

Why This Matters for 2027 Specifically

The European defence AI landscape in 2027 is at an inflection point. Budget is available. Technology is mature. Political will is — for the first time in years — present. The constraint is coordination: the institutional agreements, procurement frameworks, and trust relationships required to move from capability to deployment. Those agreements do not get made on stage. They get made in the margins.

The AI in Defence Summit 2027 is a curated, identity-verified event. Applications are reviewed. Capacity is limited. Apply at aidefencesummit.eu.