Disclaimer: I’m a paid customer of Infomaniak’s KSuite (a Swiss ‘ethical’ cloud offering). Next to the cloud storage and email and a few other extra, I recently noticed they have a ‘privacy respecting’ AI called Euria (which they also claims to be green-ish): https://www.infomaniak.com/en/euria
Overall, I’m very satisfied with their services but I’m also not much into using AI and don’t plan to change that, and I was wondering: how does one make sure an AI is indeed privacy-respecting? I mean, is there an independent audit of some sort like there are for VPNs?
You can’t. The cloud is a black box, and LLMs are not E2E encrypted. Also VPN audits can be fake. Download your own LLM if you want but you will suffer unless you have a powerful machine.
Thx. That’s what I was thinking. More or less. But I won’t download my own local AI either as I really don’t wish to use one. I was just curious to know if there was any way to… control this kind of claims.
if you can’t host it offline, it’s not privacy respecting.
I understand, thx.
deleted by creator
Offline ollama instance + alpaca. Then use a model that can also use the terminal or look things up on the web, etc.
Just because it’s an offline instance doesn’t mean it’s privacy.
privacy respecting’ AI called Euria (which they also claims to be green-ish
