Yesterday's post was about gatekeeper interoperability: WhatsApp's DMA-mandated third-party chats — a door in somebody else's wall. BEAM is the opposite pole of the same argument. A state stopped asking a foreign platform for permission and ran the protocol itself. The interesting question is not whether Belgium can read its own messages — it can, and that is the point of sovereignty. The interesting question is who maintains the code Belgium now runs on, and what the answer teaches every community that forks instead of funds.

What BEAM Is

Since March, Belgian Secure Communications (BSC) — a federal public organization — has been rolling out BEAM, a secure messaging platform intended for roughly 750,000 civil servants and military members. Defence and intelligence staff plus several public administrations are already on it; the rollout extends progressively to the rest of the federal civil service and, eventually, the Regions and Communities. It replaces the informal reality of official business conducted over WhatsApp.

The properties: end-to-end encryption, hosting in Belgian datacenters, and a closed environment where only holders of government email addresses can get in — which BSC credits with blunting phishing, since the user base is known and bounded. BSC says the system spent weeks under penetration testing by ethical hackers and security firms without a successful breach. Treat that as the operator's claim, not an audit. The structural facts below don't depend on it.

Under the hood it is Matrix. Specifically, a fork of Element's open-source stack: Element X Android on mobile (upstream v26.03.2, per the fork's own README), plus Element Web, Element Call, and Synapse — according to Element co-founder Matthew Hodgson, speaking to Resilience Media.

The License Held — and That Is the Floor, Not the Ceiling

Element's stack is AGPL-3.0. The AGPL's defining clause is the one this site keeps defending: run modified code as a network service and you must publish the modifications. BSC did. The organization's public GitHub (github.com/bsc-oss) carries six repositories — beam-android, beam-ios, beam-web, beam-call, and JS/Rust SDKs — with the Android README stating plainly: "The software including BSC modifications is made available under the GNU Affero General Public License v3."

This is worth stating because AGPL rarely gets its wins reported. Without that license, a federal state could have taken the code, modified it behind closed doors, and run it on government infrastructure indefinitely, with no outside visibility into what it changed. Permissive licenses permit exactly that. AGPL made the modifications public. A government deployed a messaging platform for 750,000 people and the source of that platform, including every change the government made, is inspectable by any citizen today. That is the license doing precisely its job: auditability as a legal obligation rather than a courtesy.

But notice what the compliance consists of. The published repositories are clients and SDKs — no modified server code appears in the organization, which is consistent with running Synapse unmodified (nothing to publish). The Android release branch is a single commit: a snapshot dump, not a working tree. Nothing in the public repositories suggests fixes flowing back upstream or a contribution relationship at all. This is letter-of-the-license publication. The letter is enforceable. The spirit was never going to be.

Compliance Is Not Sustainability

The most interesting document in this story is Element's public position. Hodgson confirms BSC met its license obligations — and immediately urges Belgium to go further: buy the enterprise distribution, buy support, fund roadmap features, join the Matrix.org Foundation as a paying member. His comparison set is NATO and the UN, "who understand that investing in the health of the underlying open source project is rather important for the success of their sovereignty initiatives."

He also named the three ways these deployments go sideways: incumbent vendors still embedded in procurement; domestic-supplier programs that produce new vendor dependence; and public bodies that run open-source software without contributing to it. He singled out funding as the persistent sticking point — governments struggle to budget for replacing tools that were never formally budgeted in the first place.

Germany shows what paying looks like. ZenDiS — the agency behind the OpenDesk suite — pays the vendors behind the open-source components it deploys, Element included, for BundesMessenger and BwMessenger. Sweden's SAFOS follows the same pattern. That is procurement as sustainability: the state keeps its fork close to upstream because it is literally paying for upstream.

Belgium has so far chosen minimal compliance. The slow-burn failure mode is structural: Element X releases on a monthly cadence, a pinned snapshot drifts a little every month, and the merge burden compounds silently. The day a security release matters, the distance between the fork's pin and upstream is the distance someone has to cover — with no funding line earmarked for the trip.

Balance requires saying the pressure here is real. Hodgson reports Element now works with more than 30 governments, that "digital sovereignty has moved from an ideological goal, to a sudden concrete need," and he names the accelerants: the Signalgate exposure of sensitive US government discussions on a consumer app, followed by an AWS-linked outage that took Signal down with it. Belgium's move is not eccentric. It is the current European baseline assembling itself — Germany's federal Matrix pilot, France's Tchap, the Bundeswehr's network — and every one of those deployments depends on a small number of maintainers staying funded.

The Federation It Didn't Use

One design note, because it changes how you read the trend. BEAM is Matrix with the federation switched off, as far as anything public shows: closed environment, government email required, no external homeserver connectivity. That is a legitimate choice for a state — it is a perimeter, drawn deliberately. But call it what it is: sovereignty by exit, not sovereignty by federation. The protocol's core property — independent operators connecting through an open spec, interoperability as the norm — is the part BEAM leaves unused. A community copying this pattern gets an open-source-internals silo. For a national government with a known user base, that may be the right trade. For a fire hall or a co-op whose members also live on other servers, federation is usually the reason to choose Matrix at all.

The Rule for Anyone Forking Their Way to Sovereignty

Credit where due, precisely: a real deployment, at real scale, on open code, with the state's modifications forced into public view by license rather than goodwill. AGPL did its job. Free is still the most expensive option when nobody funds the maintenance — and the maintainer going public to say "please also pay us" is the tell.

The rules, for a community, a co-op, or a company:

  1. A fork is a starting position, not a strategy. The moment you fork, you own upstream drift. Who merges? How often? On whose budget?
  2. Fund what you fork. Support contract, foundation membership, direct feature funding, upstreamed patches — at least one. If you can't afford any of them, you can't afford the fork.
  3. Fork in the open. Publish the tree, pin the upstream version in the README, date your snapshots. BSC did this part right, and it costs nothing.
  4. Prefer contributing to forking. Half of forks are configuration, not code. A disciplined deployment with upstream-tracked config beats a private fork nobody else can read — and beats it again at every upstream release.

Sovereignty has an invoice either way. It arrives as a support contract you chose, or as a fork you maintain alone at 2 a.m. because upstream moved and nobody paid for the merge.