Report: OpenAI fired three researchers who warned about loop transformers

What the research found

MSN report that OpenAI fired three safety researchers who warned a loop (recurrent-depth) transformer would hide computation from chain-of-thought monitors; the cached text is only a short summary and the firing and warning are not independently verified here.

Claims and how well the evidence supports them

1. Three OpenAI safety researchers were fired, per the report 93%
2. Researchers warned recurrent-depth architecture hides computation in hidden states unreadable by CoT monitors 96%
3. Published Oct 14, 2026 2%
4. Cached text too thin for further claims 9%

Evidence

• msn.com https://msn.com/en-us/news/other/openai-fired-three-researchers-who-said-loop-transformer-would-blind-safety-monitors/ar-AA2eh6yP

Actionability rated none.