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. (Read More)