Comment on Why aren't non-selfhosed AI models/apps private?
acido@feddit.it 1 month agowelcome to the real world, where corporations lie.
Comment on Why aren't non-selfhosed AI models/apps private?
acido@feddit.it 1 month agowelcome to the real world, where corporations lie.
Steve@communick.news 1 month ago
Not the real world. Just your imagination.
Corporations lie for profit. Where’s the profit for Proton in keeping peoples AI queries, when they’ve been proven not keep any other data? Literally they have nothing to gain, and everything to loose.
Skepticism and pessimism aren’t the same thing. And baseless pessimism is the dark equivalent of naivety. They’re both equally ignorant.
sanzky@beehaw.org 1 month ago
there has been in the news several instances where proton has given to law enforcements information that have hold onto. in some cases regarding journalists.
their answer is always “ah, yeah. we do keep that one, but not the other data”
Steve@communick.news 1 month ago
Of course they have to keep some basic account data. And I think the last IP you logged in from. Also email data outside the BODY can’t be encrypted. That’s just how email works. So law enforcement can get all of that if they convince a Swiss court to demand it.
But no they don’t keep or turn over anything that isn’t technically required for the service to work. I don’t know what you’d expect.
sanzky@beehaw.org 1 month ago
in that particular case the people involved were identified through their recovery email which they did not hash like ‘safe’ other providers do. they have positioned themselves as safe even for activist and journalists and have failed to deliver in that account consistently.
no surprise since their CEO is a MAGA guy
acido@feddit.it 1 month ago
can you link me your source about Proton being “proven to not keep any other data”?
Steve@communick.news 1 month ago
All the various 3rd party audits can be found here.
acido@feddit.it 1 month ago
have you ever checked them?
could you point me to where any of them proves that Proton isn’t actually collecting any data?
because I am only reading about penetration tests and such, which only tell me that their software is reliable against attackers, but that is not the point of the discussion.
also it seems there is no audit for their AI assistant at the link you provided.