Modeling the impact of safety agendas — AI Alignment Forum