How the EU Commission Built Sovereign Comms at 40k Scale

How the EU Commission Built Sovereign Comms at 40k Scale

I’ve watched enough enterprise tech rollouts fail spectacularly to know that moving from proof of concept to 40,000 daily users in nine months is either heroic or reckless. The European Commission pulled it off with Matrix, and the story they told at Matrix Conference 2026 is worth unpacking, because it reveals something important about what sovereignty actually costs at scale.

The Commission didn’t start with romantic notions about open source. They started with a practical problem: they needed to control their internal communications, and they needed those communications to interoperate with other EU institutions and potentially member states. That’s not ideology. That’s infrastructure policy.

They defined sovereignty across five dimensions: resilience, vendor diversity, data control, platform ownership, and procurement leverage. That last one matters more than most people admit. When you lose the ability to walk away from a vendor, you’ve already lost sovereignty in the ways that matter operationally.

Choosing the Right Building Blocks

They landed on Matrix as an open protocol, and Element Server Suite Pro as the implementation. But here’s the move I found interesting: they didn’t host it themselves. They used sovereign cloud providers through a dedicated tender. This is the key insight that separates real sovereignty thinking from the performative kind. Sovereignty means choosing your dependencies, not eliminating them.

They integrated with EU Login, the sign-on system staff already used. New tools fail when you force new friction into established workflows. They kept it familiar.

Federation, the ability to connect separate Matrix servers, is still under assessment. That matters because it points to the real bottleneck in interoperable systems: governance takes longer than code. You can have a working protocol; building the trust structures, participation agreements, and security frameworks that make federation actually work across organizations is a different beast entirely.

What Breaks When You Scale

The deployment uncovered four major problems that pilots never would have surfaced.

User discovery across multiple Synapse homeservers failed at first. The Commission runs several servers, and colleagues on different servers couldn’t find each other. They worked with Element to fix it. This is worth dwelling on because it shows that even well-designed open protocols need someone in the room making integration decisions when real-world deployment gets messy.

They chose user-led account creation over pre-provisioning, which meant early adopters couldn’t find colleagues who hadn’t signed up yet. That created a perception problem during the critical adoption ramp. The usability hit from a design decision compounds as you scale.

Live presence indicators, the real-time “online/away” status that feels essential in consumer chat apps, started hammering the infrastructure. They turned it off and are now working with Element on a user status feature as a replacement. Sometimes the features people expect are incompatible with the operational constraints you actually face at scale.

Infrastructure had to autoscale in response to uneven demand across a three-month rollout. That’s engineering work, not just deployment work.

The Balance Between Security and Usability

Here’s where I think the Commission’s approach maps onto a larger lesson. They did most of the security hardening before rollout, but they kept obsessing over usability. Password recovery. Device verification. Account creation flows. These aren’t afterthoughts; they’re the difference between a tool people actually use and one they resent.

That tension between security and usability never really resolves. It demands continuous attention. They’re consolidating feedback from tickets, forums, and channels into a prioritized roadmap. That’s operational governance, and it’s tedious, but it’s what keeps a deployed system actually serving the people who use it.

The Federation Question

The Commission has been interested in federation for a long time. The obvious partners are other EU institutions, but the actual ambition is broader: member states, governments, NGOs. No decision yet. They’re taking it methodically. First, the technical question: what can the technology actually do? Then governance: who, on what terms, with what scope, with what security model?

That sequencing reveals something. Technology answers come first because writing policies that technology can’t deliver is pointless. But governance takes longer to agree on than implementation usually does.

What This Means for the Industry

The Commission’s framing of three operational lessons is worth stealing. Sovereignty is operational, not theoretical. Adoption depends on usability and discoverability, not features. Interoperability needs governance, not just open protocols.

They’ve also demonstrated that collaborating with trustworthy service providers is better than building everything yourself. Element’s role in helping them scale wasn’t about open source purity; it was about having people who understand the problem space helping you solve it.

What we’re really watching is a blueprint for how large organizations can reclaim some operational control over their infrastructure without building everything from scratch, and the answer turns out to be messier, slower, and more dependent on human trust structures than the technologists usually want to admit.

Read Next