Comment on 'I had to RUN to my Mac mini like I was defusing a bomb': OpenClaw AI chose to 'speedrun' deleting Meta AI safety director's inbox due to a 'rookie error'
LLMs will often respond in a reconciliatory or obsequious manner when presented with confrontational input.