OpenAI updates Model Spec to emphasize five core values for more capable AI

KEY POINTS

1. OpenAI released an updated Model Spec that outlines five core values aimed at guiding more capable AI toward sound judgment, preserving human agency, and staying faithful to human intent.
2. The update frames these values as principles to inform development and behavior of increasingly capable models.
3. The document signals OpenAI's internal norms for aligning more capable models and could influence developer practices, model evaluation, and public policy discussions about advanced AI.