Three partners, three Member States, one evidence base

The composition rule for a European Defence Fund action is one of the most quoted requirements in the programme and one of the most often misquoted. The usual shorthand is "three independent entities from three Member States". That sentence contains two errors and omits two exceptions.

It is worth getting right, because consortium composition is decided early, changes are expensive, and the rule is not where most people look for it.

It is Article 10, not Article 9

Article 9 of the EDF Regulation governs eligible entities: where an organisation is established and who controls it. Article 10 governs eligible actions: what an action has to look like to be funded.

Consortium composition is a property of the action, so it sits in Article 10(4). Citing Article 9 for it is a small error that signals a larger one, which is that the two tests are being run together when they are separate.

The rule has two limbs

Article 10(4) requires an action to involve at least three eligible entities established in at least three Member States or associated countries.

And then the second limb, which the shorthand loses entirely: of those, at least three entities established in at least two Member States or associated countries must not be under common control, and must not control each other, for the entire period in which the action is carried out.

The word "independent" in the popular version is doing all of that work silently. Spelled out, it means three subsidiaries of one group, established in three different Member States, do not satisfy Article 10(4). Neither do two group companies plus a partner they jointly control. The second limb exists precisely to prevent a single corporate structure presenting itself as a European collaboration.

Note also the shape of it: three states for the first limb, two for the second. Those are different numbers doing different jobs, and collapsing them is the other half of the shorthand's inaccuracy.

The requirement that this holds for as long as the action runs also deserves attention: the Regulation's own words are "during the entire period in which the action is carried out". Composition is not only a test at submission. An acquisition midway through a project can put a consortium out of compliance with the rule it was funded under. The Regulation legislates for that rather than leaving it to chance. Under Article 9(7), an entity must inform the Commission of a change that might put the eligibility criteria in question, and the Commission then assesses whether they are still met and what the change means for the funding.

Two exceptions, and they disapply rather than relax

Article 10(5) provides that paragraph 4 shall not apply to actions relating to disruptive technologies for defence, or to activities referred to in Article 10(3)(c), which are studies, such as feasibility studies.

Two things about this are commonly got wrong.

Studies are usually omitted. Most summaries mention the disruptive-technology exception and stop. Studies are a second, separate category and they matter to anybody scoping a feasibility phase.

The rule is disapplied, not relaxed. You will often read that disruptive-technology calls "relax the requirement to two entities from two countries". That two-from-two figure does not come from the Regulation. It comes from work programmes, which are free to set their own conditions where Article 10(4) does not apply.

And work programmes are not consistent about it. EDF-2026-LS-DIS-RA-SMERO, a 2026 disruptive-technology call, still requires a minimum of three independent applicants from three different eligible countries. So the general rule will not tell you what your call requires. Read the call fiche.

The part nobody budgets for

Suppose your composition is sound. You now have at least three organisations, in at least three countries, that have to evidence compliance as one.

Each arrives with its own quality management system, its own documentation conventions, its own internal auditors, its own national requirements and its own idea of what "controlled document" means. None of that is wrong individually. It simply does not aggregate.

Nobody budgets for harmonising it, so it lands on the coordinator, consumes management effort intended for engineering, and surfaces at the worst possible moment: at a review, at a technology readiness gate, or when a national authority asks a question that nobody prepared for.

The instinct is to solve it at the end, by collecting a compliance pack from each partner and stapling them together. That produces a document set that is internally inconsistent, out of date the moment it is assembled, and impossible to keep current for the duration of the action.

The alternative is one evidence base, built once, in which each requirement is traced to the thing that satisfies it and each partner maintains their own part of it continuously. A control captured once can then be mapped to each standard that calls for it, and to the national requirement behind them, rather than being rewritten three times in three formats.

That is what requirements management and document control do, and it is the same discipline behind the standards each partner is likely to hold already. The difference in a consortium is only that the evidence has to survive being looked at by people from three organisations who did not write it.

Checking this yourself

• Regulation (EU) 2021/697, Article 10(4) and 10(5), and Article 10(3)(c) for what counts as a study
• The call fiche for your call, which sets the actual composition requirement where Article 10(4) is disapplied
• Article 9 separately, for the eligibility of each entity in the consortium, covered in EDF ownership and control

Three partners in three Member States is one evidence base, not three. See how ComplyTrain does that.

This describes what the published rules say as at 8 September 2026, and is not legal advice.