On October 21, at the Matrix Conference in Malmö, five public-sector deployments federated a live conversation across national borders. The participants, per Element's writeup: the European Commission, France's DINUM with Tchap, Sweden's Försäkringskassan with SAFOS Chatt, NATO's Allied Command Transformation, and the German state of Schleswig-Holstein. Three countries, two international institutions, five independently run stacks, one conversation. Element, which sponsors the conference's public-sector track, supplies three of the five stacks, and whose founders created the Matrix protocol, calls it "the world's first cross-border government federation demo." The framing on the page itself is a Bell-and-Watson telephone analogy. This site is skeptical of that genre — breakthrough framing is usually doing the persuading because the evidence is thin. Here the evidence is not thin, but it is narrower than the framing: Computerworld's report — the most detailed non-vendor account — calls it what it is. A proof of concept. Element's own text says the same: demo environments "hosted on the same infrastructure as their actual deployments," not a production cross-border network. Nobody's crisis channel carried this traffic. What was demonstrated is a protocol property — five sovereign deployments, different vendors, different build styles, federating on demand. That is genuinely worth attention, because inter-agency communication across incompatible chat systems is the daily pain of exactly the organizations this site writes for. And the most valuable artifact of the whole exercise is not the demo itself. It is the table of security postures Element published next to it.

The table is the artifact

Element's post names the people, the products, and the configurations. Tchap was developed in-house by France from community Synapse and is part of the LaSuite ecosystem. Försäkringskassan and NATO run forks of the Element app they built themselves — SAFOS Chatt and NI²CE Messenger respectively, with FluffyChat standing in for NI²CE in the demo environment. The Commission and Schleswig-Holstein run Element as it comes on Element Server Suite Pro; Försäkringskassan's SAFOS sits on ESS Classic. Five deployments is a small number, but it spans the realistic range of how a public-sector organization actually builds this: from vendor stack as-is to in-house development on open source.

The table Element published:

| Deployment | Backend | Network filtering | Enterprise control | | --- | --- | --- | --- | | DINUM | In-house from community Synapse | IP-based | Secure Border Gateway developed in-house | | European Commission | ESS Pro | Open | None | | NATO | ESS Pro | Open | None | | Försäkringskassan | ESS Classic | IP-based + geo-block | Only MDM-enrolled devices, file transfer blocked | | Schleswig-Holstein | ESS Pro | Open | None |

Read that right-hand column again. Three of the five demo edges — including the European Commission's and NATO's — ran open filtering with no enterprise controls. One locked enrollment to managed devices and blocked file transfer. One builds its own border gateway. The demo federated all five anyway, in the same conversation. Computerworld flags exactly this asymmetry as the operative fact: federation joins the servers, but it does not make their security postures equal.

There are two things a public-sector reader should take from that.

First, in the model these deployments actually intend to run — private, closed federations, where only explicitly approved servers connect and nothing is exposed to the public Matrix network — the filtering layer is the border. Element describes the intended control dimensions cleanly: who (only named, trusted servers), what (policies on which content and features cross), how (end-to-end encryption, each party keeping its own data, logs, and audit trail on its own infrastructure). That is the same design logic France's local governments chose for their closed federation — trust levels and named peers instead of a public directory, and an interconnection plan that reaches the state's systems without merging with them. The table shows what the demo actually assembled under that model: a heterogeneous mesh where your sovereignty posture is bounded by your acceptance decisions, not by the protocol.

Second, the table itself is the copyable move. Government deployments almost never publish their filtering and control postures side by side; the usual public artifact is a press release. Element's demo table is a falsifiable artifact in the same way AWS's dated disconnect drill is: it commits a claim to a form that can be checked, criticized, and required of someone else. Any organization running a federation — a city IT department, a county emergency-services network, a co-op — should be able to produce this table for its own deployment this afternoon. If you can't state, per deployment: backend, who can federate in, what filtering exists at the edge, what enterprise controls gate enrollment — you don't have a sovereignty posture, you have a hope.

The ecosystem numbers are arithmetic, not census

Before adopting the demo's momentum, price the numbers around it correctly. On the same page, Element claims 200M+ people using Matrix-based communications globally, 35+ governments with deployed sovereign systems, and "more than 70 providers" — which its own stats box breaks down as 40+ competitive vendors plus 30+ service providers. At Wednesday's keynote, Computerworld reports that Matthew Hodgson cited 65 "national- and international-scale" deployments tracked in Europe. None of these figures is audited, and the most precise one is vendor-tracked by a vendor. Treat them as evidence that an ecosystem and a market exist — they are enough for that — not as market share. The one number in this story that comes from a government's own telemetry, Tchap's more than 400,000 monthly civil-servant users on France's own stats page, is the one that behaves like a measurement.

The pattern that produced this demo started with a dead product

The Swedish strand of the demo is the one with organizational depth, because it answers a question the demo doesn't: how does a federation move from pilot to program?

Five years ago, most of Sweden's eSam member agencies — a collaboration program spanning 42 government agencies — used Skype for Business, and cross-agency chat worked because everyone was on the same product. When Microsoft deprecated it, agencies scattered onto whatever they chose: Teams for some, plus Element, Rocket.Chat, Mattermost, Pexip, Jitsi. The interoperability that had been a vendor accident evaporated, and it turned out that the replacement for a shared platform wasn't a better shared platform — it was a protocol. Kenneth Edwall, IT architect at Försäkringskassan, told Computerworld what the gap costs in practice: when agencies' systems can't talk directly, staff end up with guest accounts on each other's platforms, and those guest accounts become exactly the access-management problem — provisioning, reviewing, offboarding — that small organizations know intimately, multiplied across thirty-plus agencies.

eSam's dSam4 work group spent 2025 analyzing the options, and in January published the result as an actual document: ES2025-20, "Common federation protocol for chat in the public sector", with appendices A through J. Its recommendations are procurement-shaped, not hype-shaped: prioritize open standards in procurement; invest in expertise around federated systems; participate in standardization; run TCO analyses that weigh strategic benefits and lock-in risks; establish policies and guidelines for federated chat services.

Then the program did the thing this site keeps asking for: it built a small one first. On May 22, 2026, federation went live between Försäkringskassan on Element and Trafikverket on Rocket.Chat — two agencies, two vendors, one open protocol, each keeping its own system. eSam's announcement is clear about what the pilot is: a test that confirmed different systems can interoperate while agencies keep their own solutions, with the steering group stating its intent to recommend Matrix as the common standard for secure, interoperable communication between Swedish authorities. The honest edges are in the reporting too: Computerworld notes the pilot met most of the project's requirements with some capabilities still needing improvement, the final report lands in spring 2027, and SAFOS Chatt currently serves 19 Swedish public authorities. Peter Haglind, Försäkringskassan's IT director, made the argument in one sentence: "the way email works between organizations today" — chat is being rebuilt as the same kind of commons. The Malmö demo took the domestic pattern across a border: as Edwall put it, "it's only a few firewalls away to France and Germany, so it's the same approach, the same philosophy."

The Teams question keeps its honest answer

The most useful part of the Malmö conference is that nobody presenting pretended otherwise: sovereign chat is arriving as a complement to Teams, not a replacement, and the people who run the rollouts said so on the record.

The Commission's deployment — Element on its own infrastructure, desktop rollout to Commission laptops completed in September 2026, mobile planned, target of secure internal messaging for more than 40,000 employees — exists for resilience, business continuity, and, as Diana Gheorghe, who runs the program, put it, data ownership and "contractual leverage" in procurement — an at-times-overlooked sovereignty facet, and the one this site has been pushing since the control-layer argument. The same presentation listed the friction: employees struggled to find colleagues across different servers, the live-presence feature was disabled after performance problems, and there is "no one-to-one parity" with Teams. Politico's piece on internal reviews of the Commission's Element rollout was unflattering, and Gheorghe's response was to manage expectations, not to promise parity. On cross-border federation, her line was the opposite of a launch announcement: "Federation, for us, is an unwritten story… At this point there's no firm decision" — with GDPR among the open considerations.

Forrester's Dario Maisto provided the sober frame: the pull is real because the dependency risk is material — an organization can wake up to find Outlook or Teams unavailable, and, as the ICC prosecutor's suspended Microsoft account showed last year, it has happened already. But the shift is gradual: PoC, then production alongside, then maybe migration — and the arithmetic is brutal for anyone still amortizing a pandemic-era communications migration ("if I'm still paying for that, I'm not starting a €10 million project anew"). His sharpest point cuts the deepest: if the rest of the office runs the Microsoft productivity suite, one sovereign chat tool on the side changes less than the sovereignty slide suggests. Communications is the right place to start, he said — "just be mindful that if you start there, there might be a lot more to consider very soon."

Edwall's justification for Sweden's agencies is the cleanest version of the trade: "The idea is not that these solutions need to be exactly like Microsoft Teams. That's not the driver — it's the sovereignty part. Then we can live with not having the latest, coolest feature." The user-benefit he names is not freedom in the abstract; it is the removal of digital friction — the tool-two dance where you can't remember what you're allowed to put where.

Two structural facts close this section. First, the bridge shortcut into the incumbent estate is narrowing: Element no longer offers its Teams and Slack bridges to new customers, and bridging is feature-limited anyway. Second, protocol-level interoperability is also arriving from the other direction — the law — which is a reminder that the standard-vs-silo question is now being decided in more forums than any one vendor's roadmap.

What a five-party demo does not answer

Keep the demo's claims inside the demo's scope. None of what follows is settled by a proof of concept on conference day:

  • Key custody across borders. Matrix federation supports E2EE, but who holds the keys when a conversation spans two jurisdictions — and which jurisdiction's law reaches the key material — is a legal architecture question, not a protocol one.
  • Identity and enrollment asymmetry. One participant in the demo allowed MDM-enrolled devices only; three allowed all devices. A federated conversation inherits the most open posture in the room unless policy intercepts at the gateway. The filtering table is per-server; the risk surface is per-conversation.
  • Retention, audit, and log law. Each party keeps its own logs, so each party's logs obey its own rules. That is the right outcome for sovereignty and a genuinely hard one for cross-border investigations and records-retention obligations.
  • Load and failure behavior. The demo ran on conference infrastructure in a controlled window. Crisis traffic — the moment when inter-agency messaging actually matters most — behaves differently, and nobody's production incident channel was stress-tested.
  • Governance. Edwall's own list of open questions stands as the agenda: who should interoperate with each other, how open should the network be, and who is allowed in. Those are organizational constitutions, and no protocol demo writes them.

Watch two concrete things next: eSam's federation final report in spring 2027, which will grade the two-vendor pilot in detail, and the Commission's federation feasibility work, where the "unwritten story" either acquires a decision or doesn't. Conference recordings are already up on media.ccc.de for anyone who wants the primary material instead of this and vendor posts.

What to steal

  • Write interop into the contract. eSam's first recommendation — prioritize open standards in procurement — is now backed by a public existence proof: five government deployments from three countries and two institutions, different vendors, one conversation. "Must interoperate with Matrix-compatible deployments specified in Annex N" is a requestable line item, and the demo is the citation for why.
  • Gate the federation, name the peers. Closed-by-default with named servers and a written policy for who gets a door — France's local-government federation and Tchap both run this way, and the demo's secure-border-gateway model is the same idea. Copy the gate, not an open directory.
  • Publish your filtering table. Backend, edge filtering, enterprise controls, per deployment. It took a vendor one blog post to make five governments' postures legible; your stack should not be less legible than NATO's demo environment. This is the same genre move as the AWS drill: an artifact that can be checked, which is the only kind of sovereignty claim that means anything. Two artifacts in one month — a dated disconnect and a posture table — is the genre shifting from adjectives to evidence.
  • Price the guest accounts you already run. The non-interop cost never appears as a line item called "interoperability"; it appears as guest-account sprawl, access reviews, and the offboarding risk Edwall described. Count them before deciding interop is not worth a budget line.
  • Set the expectation on purpose. A sovereign messaging rollout succeeds when leadership says plainly what Försäkringskassan and the Commission both said: this is not Teams, it is not meant to be, and it exists so the organization can choose its own infrastructure and know what channel carries sensitive work. "This is not an IT problem, this is a board-level decision," as Forrester put it.

Sources: Element: the cross-border federation demo (Oct 21, 2026); Computerworld: Matrix gains ground in Europe's public sector (Oct 23, 2026); eSam: federation go-live announcement (May 22, 2026, in Swedish); eSam: ES2025-20 Common federation protocol for chat in the public sector (Jan 30, 2026); eSam: Sweden goes live — Element companion post (May 26, 2026); The Matrix Conference 2026 recordings; France stats: Tchap monthly users.