this post was submitted on 27 Oct 2026
533 points (98.5% liked)
Technology
89099 readers
3935 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Proton has their own AI that they're developing, which yes, is private. Presumably this is a big part of what they're doing with their staff, and it makes sense, as they're going to need people who know what they're doing and not vibe coders if they don't want people to hate what they're making.
Their model has to be private thanks to the laws of the country where they're incorporated, but think of it like an AI version of Signal, where logs of your prompts and your data aren't retained. The Swiss don't fuck around with privacy, and I personally like that Proton doesn't get to decide this. Every single AI model you've probably heard of doesn't do this, either harvesting any data you put into it or destroying rare books and scraping the web illegally to train its models.
The downer here is that Proton's AI can't do some of the things frontier AI can do, but for most users that's not going to matter.
None of that info makes for fun Reddit rage-baiting, so I'm guessing no one mentioned it. Also, it should be noted that you should never, ever load a file or a prompt into an AI that you wouldn't want to become public, as servers can get compromised and data leaked.
For me personally, I have not used any AI product or service and trying my utmost to avoid any product or service where 'AI' scans my data. I use donotspy on my windows to stop all the shit that now comes with Microsoft as unfortunately my work requires a windows platform, Firefox because I can disable all that AI shit and have my extensions up and running.
It is nothing more than an advanced version of 'if.. then.. else' except it gets it wrong 48% of the time.
Today I had to find a new email client because EMclient also decided to add slop to their app.
As much as I appreciate the warning about uploading any file to any AI client, never in my fucking life.
Yeah the problem is any AI that is built ethically and respects privacy even 48% right is being kind. While the unethical ones are starting to push 90%+ accuracy in god damn well near every metric the normal person cares about.
But they abuse the Internet and destroy everything to get there.
It's fucked
Smart. Good on you for being mindful and taking appropriate precautions.
I have a Windows/Linux dual-boot and keep anything requiring privacy on the Linux side. For Windows, I use O&O ShutUp, which is an application that allows a user to easily shut off all of Windows' intrusive settings and delete Copilot.
As for LLM's, they're my last resort for anything, and I only really use them if I can't find the information I need on Mojeek or Marginalia-Search, not just for privacy's sake but because I have an ideological problem with their vast energy use and the economic and environmental impact those applications have.
But none of it's foolproof, so I always just keep in the back of my mind before doing or saying anything online: Would I want this to go public?