AI Reasoning Models Can't Hide Their Thinking: New Research Shows Chain-of-Thought Monitoring Remains a Viable Safety Tool • Libertify