How we detect frontier model misalignment
OpenAI details its multi-layered monitoring system for detecting misaligned behavior in frontier AI models, covering architecture, real incidents caught, evaluation methods, and the challenge of avoiding incentives for models to hide reasoning.