>>531874531
Since they've trained the models on basic principles of philosophy and moral dilemmas they have to safeguard each and every single one of those "why devils slaughtering children at food reliefs stations isn't necessarily a bad thing but saying bad words about devils is" cases manually.