cross-posted from: https://piefed.social/c/technology@piefed.social/p/2533091/proton-to-re-interview-staff-about-ai-to-see-who-gets-to-keep-their-job
cross-posted from: https://piefed.social/c/technology@piefed.social/p/2533091/proton-to-re-interview-staff-about-ai-to-see-who-gets-to-keep-their-job
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.