OpenAI has released an updated version of its Model Spec. This revised framework outlines five core values intended to direct artificial intelligence systems as they grow more capable. According to the release, the initiative aims to guide these advancing models toward sound judgment.
In addition to sound judgment, the framework focuses directly on preserving human agency. It also seeks to ensure faithfulness to human intent across model interactions. Through these five core values, the documented rules establish how OpenAI plans to steer capable systems as their capacities expand.