Researchers present ReasoningShield, a benchmark and safety moderation tool for Chain-of-Thought (CoT) reasoning in Large Reasoning Models. The tool and benchmark aim to address safety challenges in CoT moderation, outperforming existing tools by up to 35.6%.