I have always assumed that everything that I type into an LLM is guaranteed to be completely insecure, and entirely public. I would never, in a million years, put anything sensitive into an LLM, any account and password is unique.
I mean, I figure whatever I type is going to be used as training data for the next iteration, and I assume anything that goes in, is retrievable to some degree.
these are not two sides. The system is working as some intend so needs to be dismantled, at least large parts of it, to fix it.
People always forget, nuance exists.