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