Model Misalignment
OpenAI details layered monitoring for frontier model misalignment
OpenAI said it uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior before incidents escalate.