Safer Reasoning Traces: Measuring and Mitigating Chain-of-Thought Leakage in LLMs | ScienceToStartup | ScienceToStartup