Source OpenAI · Published · Open the original ↗
Lab notes
How we detect frontier model misalignment
OpenAI uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior so responders can intervene before incidents escalate.
OpenAI publishes this post on its own site. Supence shows only what the publisher allows to be quoted.