No, it's not, there is a lot of control You clearly don't know much about it The AI handle float point values, 0 to 1 Then it's translated to words using a corpus At any point you can say "don't answer the question if it contains these words "murder, kill, suicide"