20 sats \ 1 reply \ @OriginalSize 7 Feb \ on: AI terminal and security devs
In my interactions with LLMs, copy-paste has been a useful step for thinking twice and doing sanity checks.
I'd only run this in a VM to start. Then I'd want to see and approve what's sent and received before execution.
With great power comes great responsibility.
this makes sense. preventing network access to it for a while would probably also work. there shouldn't be any reason it's sending data out unless i explicitly ask for chatgpt help.
reply