I asked 84 strangers if they care about AI privacy. Here's what actually surprised me.

A few days ago I posted about why I built ArcanAI (arcanai.co) , a privacy layer that anonymizes your documents before any AI model sees them. 84 comments later, I learned more than in the weeks before.
Here's what surprised me.
Nobody thinks they're doing something wrong. They paste, get the result, move on. The risk accumulates invisibly through repeated successful use.
The breach framing is too narrow. The deeper issue is that AI platforms are built to receive as much data as possible. Your data today shapes what gets sold and recommended tomorrow. The oversharing isn't a bug, it's what serves the platform.
And the real buyer isn't the person pasting the document. It's the compliance officer responsible for what the whole team does.
One question for this community: when you use AI, how often do you upload documents that contain sensitive data, client names, financials, personal info? And do you think about it when you do?