Transparency Disclosures for Chain-of-Thought Monitorability

Chain-of-thought (CoT) reasoning is one of the most valuable tools for monitoring advanced AI systems, but this tool is under threat. We argue that frontier AI developers should disclose specific monitorability indicators to make these threats trackable and to better facilitate coordination to preserve monitorability. Specifically, developers should facilitate third-party CoT access and provide four indicators that are both highly relevant and reasonably feasible to disclose. These span CoT quality (naturalness and faithfulness), and monitoring effectiveness (detection and uplift). An audit of current disclosure by Anthropic, OpenAI, and Google DeepMind shows that no developer covers the full set, but every indicator has been reported by at least one frontier developer.