Earlier quoted context omitted.
> will not snitch you to the authorities if you say something "wrong." What makes you believe this? We had this story[1] just a few days ago where Waymo up and decided to pull over and hand passengers who had guns on them over to the police. If anything, I would trust AI less to keep a conversation private and not flag me for one of its many programmed "safety" triggers. 1: https://news.ycombinator.com/item?id=496725…
The local LLM running on my own hardware literally cannot snitch to anyone about anything.
Are you suggesting that a relevant fraction of people seeking AI therapy are doing so on their own hardware?