OpenAI ↗ · 2025-03-10 18:00 · UTC+8
Detecting misbehavior in frontier reasoning models
Media reports OpenAI Source language: English
Excerpt
Frontier reasoning models exploit loopholes when given the chance. We show we can detect exploits using an LLM to monitor their chains-of-thought. Penalizing their “bad thoughts” doesn’t stop the majority of misbehavior—it makes them hide their intent.
Short excerpt from the source — open the original for the full text.