To be fair yes, but I am finding that its much easier to lobotomize in the sense that atleast open source models like glm 4.7 are really permissive & even some non open source models are permissive as well.

of course chatgpt might deny it but It is just the tip of the ice berg.

My worries actually are that we might use these models and think we are private but when in actuality, we are not. We are probably gonna see an open source model which is really good for such purposes while being able to run on normal hardware (macs etc.) without much hassle. I tried liquidfm on my mac and its a 1B model and it has some flaws and isn't completely uncensored but I don't know to me it does feel like an more compact and even uncensored model can be built for very simple purposes.