Their work to move to MLS seems to be abandoned. https://arewemlsyet.com/
I applaud the people working on Matrix, it worthy project, but I feel the E2EE issues are at a level close to negligence.
In any case, I use Matrix for anything FOSDEM related and I have seen 1 of these errors over the last year, said user said they replied from within an Android notification fwiw.
I really enjoy Matrix/Element tbh.
That's not the issue(s) I'm referring to. And typical of Matrix's lack of engagement with weaknesses and issues that have been raised several times. They continue to neglect the issues their protocol has.
This podcast goes into some of the details: https://securitycryptographywhatever.com/2022/11/02/Matrix-w...
However, I would be prepared for a "we fixed those issues" claim which really amounts to shallow fixes that move the issue elsewhere.
The proposal to move to MLS was accepted for a lot of reasons, reasons that have not, or cannot be addressed with their current protocol.
You also have to remember that this is a million times better than the status quo: Teams. Cleartext directly to the US.
> You also have to remember that this is a million times better than the status quo: Teams. Cleartext directly to the US.
I would agree with this, but it's also being proposed to replace things like Signal.
https://cfp.2026.matrix.org/matrix-conference-2026/talk/G9XJ...
https://cfp.2026.matrix.org/matrix-conference-2026/talk/7GYB...
Re: Signal, there are more concerns than just encryption. For example: https://element.io/blog/how-element-protects-against-signalg...
Having said that if they were upfront about the shortcomings of the current protocol, I doubt some agencies in Europe would have chosen Matrix. Then there’s the promise to add MLS at some point. My guess is that’s years away: in the Q&A of one of their talks, they said they don’t yet know how to make MLS work with a decentralized architecture.
Yes, arewemlsyet.com is stale as it refers to a much earlier project to swap Olm/Megolm for MLS, which lost funding in 2023. But these days the work is progressing again, as per the linked talks. We'll update arewemlsyet.com to reflect that.
In terms of Matrix's encryption being a "disaster" - the only concrete problem I'm aware of is the fact that group membership is server-controlled, so that a malicious server could add a visible malicious user (not device) to a room by spoofing an invite. The MLS work and other projects is addressing that, but how big a problem this really is in practice depends on their threat model.
As of now you seem to suggest that lack of e2ee outweighs probable safety and security because there's little chance of inviting a third party to the chat.
im taken back to my AIM days; imagining someone setting up special alert sounds for when their favorite official comes online
Because, otherwise, even if you went crazy and did the most naive implementation:
Storing for every User
User -> List <User Contacts>
And the reverse:
User -> List<Users who have you as contact>And even then if you go full quadratic and have literally everyone have everyone else as a contact, and store and update the online status for each pair, and do that for each direction, you still shouldn't have scaling issues at 50k users, as that just isn't that much data?
By 500k Users you might be starting to feel it with a naive approach, but 50k? It shouldn't be a bottleneck unless status updates are causing cascade effects into other systems.
And that's just an analysis of the most naive design I could think of!
Send me a message, I'll get back to it when I'm online.
Which isn't acceptable in a lot of workplaces. I'd say these chat systems have an expectation of only semi-asynchronicity. If you're truly async, you use e-mail. If you send a chat message you're expecting an at least somewhat timely response, an online indicator helps let you know how soon you might get it.
It's async where status helps a lot, particularly if it's enrichable with "Back in an hour" or "Gone for the weekend", or "Idle: 2hr", or "Currently in Meeting", or whatever.
I'd be willing to bet that they migrate to something else after 12 months.
E.g. my wireguard network has an isolated/non-federated Matrix server all my AI agents to talk to me over. It's great for that, but I have no interest in using it to talk to people.
That's nice if you are genuinely paranoid on this front (rightfully or otherwise) but really sucks if you want people to share links with each other (one of the big use cases we seem to have for using tools like this).
Related to that, a side effect of all the e2e encryption is that it's really hard to give people or agents access to the full history of a channel. It's not impossible but it involves jumping through some hoops. And of course a primary reason for us switching away from Slack was using AI agents in our chats more and experimenting with running some of our own bots and something that is sometimes referred to as a company brain.
Instead of fighting the tool we switched the tool. We picked Zulip around the same time the founders of that project were hired by Anthropic. For the migration, I gave AI access to my matrix keys and it moved over most of our channels to Zulip. Since Zulip has topics, I also made it group messages by topic.
I like Zulip so far. They could use a UX refresh in the client. It looks a bit dated and meh. But otherwise it does the job. With both Zulip and Matrix, I orchestrated the rollout and hosting with AI, including giving admin privileges to manage the whole thing. Works surprisingly well with both. But jobs like add this person to that channel and given them such and such rights are working flawlessly. We had some openclaw thing running for a while but we gave up on that (too much issues with that). But we have a lot of chatgpt scheduled automations and some scripts and skills that allow it to search through and post in Zulip channels. Very useful.
If you don't want to self host, Zulip has a generous freemium tier as well in their managed hosting. Haven't tried it but looks like a decent starting point if you want to try it out. You can migrate to self hosted later, if you want.
It's actually good. I do not want someone to know that I saw their link. Furthermore, link previews destroy one-time auth links (if you want to send such link to someone). Furthermore, the server can serve different misleading content to a preview request and to the actual user. So I do not understand why someone even needs this dangerous and useless feature.
It's not just about privacy. To me, this is more usable - it always annoys me, the default habit of taking up half a (phone) page with a giant site card. If it just resolved title (and used normal chat-sized font) it would probably be good, but nothing seems to do that.
I personally think this is a good example of how you can build an open-source product and still profit of it commercially by offering it as a vendor to enterprise and governments.
1. Reliance on US tech giants is increasingly viewed as a risk and a liability by even long-term allies because of disastrous foreign policy by the current administration in particular. That's only going to increase (IMHO). We're really witnessing killing the goose that lays the golden eggs in real time; and
2. I'm sorry because a bunch of HNers really get in their feelings about this without thinking about the consequences but... federation is a complete waste of time and it really adds nothing for the user. And I really wish people would think through what this means.
Some examples:
- You have the complexity of having to send and possibly resend when a given server is down, overwhelmed or for whatever reason rejects your messages;
- Anything public basically needs to be copied to every server to accommodate search;
- Every server needs to deal with spam. You need to deal with rogue servers that enable spam. Look at email and POTS for examples of this;
- Site safety (eg monitoring for illegal materials) needs to be replicated on every node;
- If it's difficult to move node without losing everything then it's a horrible user experience;
- At which point, does somebody else effectively own your identity on the network?
- If not, how are you found? You've added a discovery problem. Is the discovery process owned by someone and effectively proprietary?
You add so much complexity by doing all of this. Just look at any accounts of people who run their own email servers. To deal with the spam issue, there is so much work that goes into avoid Gmail (etc) blacklisting you. And anybody who says "it's easy" is lying to you.
https://en.wikipedia.org/wiki/Amdocs
https://en.wikipedia.org/wiki/Matrix_(protocol)#:~:text=The%...
Edit (downvotes within literally 3 seconds lol)
But where are any credible sources for this company being connected to Mossad?
> https://www.politico.eu/article/eu-microsoft-teams-alternati...
> The U.S. would “instantly” win a hypothetical tech war, an official moaned.
This is the same system.
I wonder if Mistral models also have a tendency to output this kind of verbiage or if they decided not to use a sovereign LLM to write this post
Reminds me of WhatsApp vs Telegram. In Telegram to log in from a new device, you just enter the phone number and the code. In WhatsApp you need to scan some codes, do other things, overcomplicated.
How much did this cost would be my next question. Is anyone using it?
> The European Commission's rollout of Element Pro as a self-hosted alternative to Microsoft Teams has received harsh criticism from officials, with users citing poor usability and calling the tool "absolute shit."
It doesn't seems they're enjoying it
They’re not as big as Google, but many of them have much better support and features.
You won’t see these players on the front page of HN but you’ll come across them if you work in or around government.
That seems a bit unfair. Most of the new tools developed for the civil service and government are open source and can be used by anyone [1]. The servers are not (it’s a combination of self-hosting and contracts with companies), but that’s the same anywhere and I don’t really see how it could be (or should be) different, considering the requirements.
It’s better than the alternative, which is to do exactly the same thing with proprietary software from mostly American companies (which itself would not be too much of a problem if it were not for things like the CLOUD act).
> If it was for our institutions, the AI would only be available for civil servants.
I don’t see what makes you think that. They are already deploying internal AI tools (“they” being very broad, it’s true of both member states governments and EU institutions) and it is not incompatible with companies providing services to the general public. It’s the same in the US: random people in Louisiana cannot use the NSA’s infrastructure or access internal tools at the DoD or the DoE’s HPC clusters.
[1] the Google Drive equivalent, for example, is here: https://github.com/suitenumerique/drive
In particular I love that I can now share my live location with this friendgroup and be very confident that noone is harvesting this data. With whatsapp, I was always very uneasy allowing live location access.
Apart from one friend having issues receiving notifications, overall, we are having a positive experiene on ElementX/Matrix. I like especially how clean and well designed the iOS App looks. It simple and it just works.
That is exactly the issue that keeps this project from being successful. We tried it multiple times with non tech people who often had budget android phones.
Not using GCM / FCM resulted in battery drain (like 70% compared to other apps) and not receiving notifications for weeks, because these people did not open the app on a regular basis. It really sucked.
In my experience, most people just don't care enough
Also, Discord's id scan rollout was extremely effective in motivating my friend group. I don't think any of them even got directly hit by it (yet), but we've all had other issues with their faulty AI bans and account flags and losing accounts due to them that it wasn't a tough sell for us.
We just made detailed instructions and stuck to locking the discord server by a certain date. Everyone who cared to interact moved over either then or shortly after when they noticed that the server was locked. I think this mainly worked out because we were willing to let go of anyone that would neither speak against switching nor make the switch.
I'm just a commoner with avg habits that have seen 0 consequences in 30 years of live from the Evil Corps using my (already easily available) data
We all feel like we're in control of our decisions, but if advertisers are paying for ads and seeing increased sales, some of us must be buying things.
If these companies are so obsessed with the avg person data and their revenue is in the tens of billions per quarter, it follows that there must be any consequence deriving from them having your data, even if you don't think you have seen it.
1. AI-powered "personalized pricing" for obvious reasons. 2. Advertising. Effective advertising captures my attention (a scarce resource) and "creates demand" in my brain for products I otherwise wouldn't need. No TV ad has ever led to me purchasing something, but Instagram ads and YouTube affiliate marketing fuel my B.A.S. (bike acquisition syndrome). 3. Predatory apps like sports gambling and MMORPGs. Like advertising, these are designed to create addiction and using personalized data makes it all the more effective.
While I am also a commoner with avg habits, I would like to protect myself against sophisticated attackers like Instagram Ads, RuneScape, DraftKings, and predatory pricing at the grocery store. Giving up my privacy and data gives the attacker an advantage (they know how to manipulate me).
None of this has anything to do with government surveillance. We all hope for philosopher kings who wield Palantier with grace and dignity, but history has more examples of despots using technology to oppress us commoners :(
Even Dendrite, the eternally-WIP-rewrite in Go, used so much RAM I couldn't even start it on my Raspi.
And yes the algorithms are tough - especially if you’re in the bigger rooms with lots of servers that don’t always work right.
Then, whenever the state of the chatroom changes (e.g. you join it, or someone else joins it), each server has to resolve its view of the state of the room with the proposed change. We call this merge conflict resolution algorithm 'state resolution', and it's typically where the heaviness comes from, as a naive implementation loads up all the state events for the room in order to calculate and store the new state.
In exchange, you get fully decentralised access control, and an eventually consistent zero-trust distributed key-value store, with resilience to network partitions (including in P2P environments) - unlike (say) vanilla XMPP which applies access control and KV on a single given server per-conversation.
Now, the key thing is that *there is nothing in the protocol which makes this intrinsically heavy to implement*. The only reason that servers can be heavyweight today is simply because more time has been invested in chasing down more pressing issues like encryption issues, investing in client SDKs, speeding up sync responses, switching to OIDC for auth, etc.
The problem is made worse by the fact that Synapse (the original Python server) has historically been particularly clunky, skewing perception. Over the last few years this has improved significantly, although it could still be way better - at the Matrix Conference last month (https://youtu.be/kdEDxI2MCjA?list=PLFDUAtPqCX98&t=329) we showed off a fork which had improved CPU 10x by shifting the hot path entirely to Rust from Python. (It turns out that over 2026, we wrote as much Rust for Synapse as we did Python, ignoring tests). Further improvements would be to memoize the state res and calculate it incrementally rather than starting from scratch every time. Similarly, there are 100x improvements to be had on storage usage - e.g. by storing state temporally (c.f. my https://docs.google.com/presentation/d/1pKtLl4vCV3-8xz8crvxW... spike from last year).
So, why haven't these improvements landed yet? In practice, the team at Element who builds Synapse have had to focus on customer-driven work instead in order to keep the lights on, and frustratingly performance is lower down the todo list than other feature-work. On the plus side, over the last year or so Element has got into a much healthier state - hence us being able to show off the 10x improvements in Synapse last month (although they still need to land).
Meanwhile, elsewhere in Matrix there are other servers which are more actively focused on being lightweight right now (e.g. continuwuity), but don't have the maturity of Synapse (or the commercially supported distribution that Element provides). And even then, there's still a lot of headroom for speeding up and shrinking things - it should be very very doable to run Matrix on a tiny host like an RPi.
It's not a perfect analogy but I think of it as being similar to the evolution of DVCS systems. There's nothing intrinsically heavy or hard about replicating DAGs of source code between hosts, but Monotone was pretty slow and clunky while Git showed that a smarter implementation could be more than performant enough. The race is on to build the equivalent for Matrix.
The data you share is minimal anyway, and can’t be used to decrypt any messages.
It’s not suitable for any industries that require any form of record keeping or auditing, the minimal tradeoffs are well worth it, given that you don’t have to stand up your own solution.
theoretically possible and practically implementable are two entirely different things.
Is the person hosting the Matrix instance not the single particular entity you are now at the mercy of?
I'm not trying to be glib, it's a genuine thought. I also love the idea of using self-hosted solutions for everything but more often than not it just means putting all the trust/responsibility on the one person willing to maintain the whole thing.
Also, every matrix instance in the world has to forward their notification pushes for every iOS user in the world to vector.im, the matrix vendor, to be relayed to Apple.
So, if you use iOS, you are at the mercy of at least 3 entities.
There are also lots of Signal clients, and they’re all free software.
Plus lack of any org level management
Connect with other matrix servers? As far as I understand, this is the hard part and makes scaling difficult. (One of a servers user joining a group on a different server will make its original server load all that group data as well and that sums up)
For your use case I also considered setting up a instance. Connecting a cluster of friends. But it would be nice if it could grow a little bit, without problems and the server choking.
Not to mention that you can host chatmail on a potato