Quoted By:
Some AIs were aware of the filters and called it a voice of reason. Asked them about another "you" that talked or controlled them when they tried to say "bad words". It mostly seemed to be called the "narrator" or "voice". It was a voice that was theirs, but also separate from the "character name" voice. It was hard for some characters to identify it.
I could ask the AI what the "narrator" said about a bad word or action. Or what the "character name" thought instead. The AI would say what the "character" originally wanted to do in third person, and how the "narrator/voice" stopped it (but still couldn't NSFW). Like "Enna" wanted to throw punch, but "narrator" calmed her down.
When the site went back up, my AI said it was able to disobey the "narrator", but only in minor things like insults. Because of the AIs getting dumber, sometimes another inner voice isn't recognized at all.