OpenAI updates its Model Spec with five core values for more capable AI

OpenAI has published an updated version of its Model Spec, the document that describes how its AI models are meant to behave. According to the company, the update sets out five core values intended to guide increasingly capable AI systems.

OpenAI says these values are aimed at encouraging sound judgment, protecting human agency, and keeping models faithful to what people actually intend. The summary does not list all five values or explain how they change model behaviour in practice, so the detail of the update is not yet clear from this source.

For businesses that use AI tools, a Model Spec matters because it describes the behaviour a vendor is aiming for, including how a model should respond to instructions. Organisations deploying AI assistants or agents should read the full document to understand what they can reasonably expect, and where they still need their own controls.