How we monitor internal coding agents for misalignment

How OpenAI uses chainofthought monitoring to study misalignment in internal coding agents—analyzing realworld deployments to detect risks and strengthen AI safety safeguards.