Language Agents Reduce the Risk of Existential Catastrophe — AI Alignment Forum