Being a Robust Agent — AI Alignment Forum