I thought demand for AI privacy products would come mainly from ideology. What I’m noticing instead is that it often appears after small moments of friction when users hesitate before typing something delete a message or decide not to ask a question at all.
Looking at @OpenGradient and OpenGradient the interesting part isn't just the privacy claim. It’s the attempt to change the mechanics underneath the interaction.
I was thinking about this messages get encrypted on your device and your identity is stripped before it even hits the model. So privacy stops being just trust our policy and becomes something you can actually verify.
Which makes me wonder do people really want privacy first AI or do they just want the hidden costs gone? The way we self censor today because someone’s watching. If the system changes and that fear disappears do we suddenly want more from AI Or was that demand always there just buried under the risk of being tracked?
I’m not sure the answer is obvious yet. What I’m watching is whether these design choices translate into different user behavior over time especially as OpenGradient and $OPG continue to build around privacy as infrastructure rather than a feature.