I am LOVING Proton’s new AI, Lumo! Are there any concerns I need to take with Proton as a company? Heard something recently about them getting fashy.
“Getting fashy” idk what kind of concerns you’re thinking.
That’s all the person said to me about it. If you’re a follower and frequent user of proton and are saying everything is alright then I’ll go with that after a bit of research.
What proton offers as a paid service is compelling. It does give you the opportunity to provide metadata that can be used against you, which is how the legitimate concern over its cooperation with law enforcement played out.
I would recommend against choosing security or privacy services based on the politics of their c-suite. In a lot of ways it’s similar to getting firearm training or equipment. You might feel like you’re materially supporting your enemy, but specifically in this particular political moment you aren’t in a position to sus out the depth or veracity of someone’s expressed politics and the point is to get the right understanding, equipment and systems, not to vote with your dollars.
Further, you want to understand how the system you’re purchasing access to works. Email, for example, cannot be private without some sort of additional cryptography. SMTP is best effort only, and even in the case that the system doesn’t fall back to unencrypted transmission, it relies on certs that are often spoofed by the end users isp as a matter of course and allow mitm attacks.
Information input to an llm will be used as a training dataset later on down the line so that means anything you use the llm for shines the panopticon beam onto you.
Well what I don’t love is that it’s not open source. The interface is open-source, and it’s definitely an important step, but the actual “AI” or LLM and the backend it uses https://proton.me/blog/lumo-security-model is not.
That’s unfortunate because other open source project that use “AI” (quotes here because the term is… it’s just not intelligence but anyway) like Immich, even though focusing on another topic (namely images) could benefit from that for https://github.com/immich-app/immich/tree/main/machine-learning
So… I’m a Proton client, Visionary for years, and I do NOT like AI… but if they absolutely mush shove it down my throat at least :
- be open source with the whole toolchain
- tell me which models are used
- tell me how the datasets for models were collected
- tell me how much energy it users
otherwise I will assume something under is not right.
Overall it feels like Proton has great intention but more and more uses the excuse of privacy to keep secret that are unfortunately detrimental to individual users and the privacy community overall.
Thank you for taking the time to explain this! It really helps me better understand it.
With pleasure, thanks for the positive feedback. I’m not an expert but I did write a little bit on self-hosting AI on my wiki https://fabien.benetou.fr/Content/SelfHostingArtificialIntelligence so feel free to have a look and if you have questions, happy to help if I can.
Awesome! That’s exactly what I started working on after your response. I got Llama chat and started working with the tiny Llama model. So I’ll definitely give that a read.
Ask it was llm it is derived from. Spoiler. It won’t tell you
The Proton CEO has said multiple times how much he admires Trump—and this is after he started displaying his fashy colors. He might care about privacy, but I can’t trust him not to sell me out if his favorite celebrities “ask nicely.”
Edit: And they just recently censored journalists “by accident”: https://theintercept.com/2025/09/12/proton-mail-journalist-accounts-suspended/
Thank you for clarifying! That’s what I was wondering about. You answered it. Much appreciated.
Also note that this is now the second time (that I’m aware of) where they’ve censored or helped to silence journalists or activists.
https://theintercept.com/2025/09/12/proton-mail-journalist-accounts-suspended/
They basically did the equivalent of a YouTube takedown and left it to the affected people to show why they shouldn’t have their accounts suspended, rather than doing the reasonable thing of making the original claimants prove why the victims should have their accounts suspended. It’s decisions and actions like these that make me wary of the company.
This is misinformation.
Alright. I leave it to you to show where I’ve made a mistake in my assessment. I’ll change my tune if you provide a compelling case.
Yeah, reinterpretation of some lines with context removed.
Wow, what a great service. Definitely a bastion of privacy.
https://theintercept.com/2025/09/12/proton-mail-journalist-accounts-suspended/
Can’t resist. What did Lumo say when you asked it about Proton’s privacy and security?
