OpenAI · Official · Blog Article
Reasoning models struggle to control their chains of thought, and that’s good
Summary
OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety safeguard.
Opens in a new tab. Subvolts summarizes and links; the full piece belongs to OpenAI News.
Where the other five stand
Related: Google: Gemma 4: Byte for byte, the most capable open models · Anthropic: When AIs act emotional
Hype check
2/5Worth a look
Rated low: routine. Worth knowing, not worth rearranging your day for.
Who's talking about it
- Deadline Day for Autonomous AI Weapons & Mass SurveillanceAI Explained · Web · Feb 27, 2026
Prior coverage our earlier items on the same thing
- Making AI work for everyone, everywhere: our approach to localizationOpenAI News
- Update to GPT-5 System Card: GPT-5.2OpenAI News
- Introducing gpt-oss-safeguardOpenAI News
- gpt-oss-safeguard technical reportOpenAI News
- SafetyKit scales risk agents with OpenAI’s most capable modelsOpenAI News
- Why language models hallucinateOpenAI News
Questions people ask
- Where can I read the full story?
- On OpenAI News. The "Read this on OpenAI News" link above opens the original in a new tab. Subvolts publishes a summary and analysis, never the full piece.
- What does this mean for ChatGPT?
- OpenAI introduces CoT-Control and finds reasoning models struggle to control their chains of thought, reinforcing monitorability as an AI safety…
More from OpenAI News 1017 more
Page generated Sep 3, 2026. Summaries are Subvolts' own; the story belongs to OpenAI News.