I was thinking the other day, that even if from Kagi, no account related data is sent to the various LLM services (via the assistant), the custom instructions we have for the assistant are sent to these agents. So, I was wondering if there is (at least a contractual) guarantee that this (possibly) unique and (mostly) constant blob of text is not used by these services to make a profile of the user. What I mean is them using said blob of text as a "user identifier" and thus being able to say "ah, this new prompt is also from the same user!" and consequently gradually add up these bits and make a "profile" of the user (or in any other way using that common blob of text to make inferences).
Would someone from Kagi be able to give an explicit and clear answer to this. Based on that how I use the assistant might change and I may also have to remove the custom instructions if there is no safety from the likes of Facebook (now Meta) who even made shadow profiles of people who never used their services.
My apologies for mis-categorising the post but I don't see a way to just say "discussion" or "question". I only see "suggestions" or "bug report".
