Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety — latest news, discussion & facts — IJR