news
Reasoning models struggle to control their chains of thought, and that’s good
March 5, 2026
OpenAI introduced CoT-Control and reported that reasoning models struggle to control their chains of thought. The finding reinforces monitorability as a safety safeguard, since weak control over internal reasoning may make those chains easier to inspect for misbehavior.
OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety safeguard.
Source: openai.com