An updated Model Spec for more capable and beneficial AI

Why it matters

OpenAI has revised its Model Spec, the public document describing how it wants its models to behave, restructuring it around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The company frames the change as necessary as models grow more capable and take more consequential actions in the world, requiring judgment in situations their instructions do not explicitly address. The revision adds guidance on context, oversight, and agency, adjusts some behavioral defaults, and retires much of the detailed style guidance from earlier versions. Faithfulness to human intent serves as the foundation, with OpenAI setting boundaries, developers directing applications, and users directing the model's work within those constraints. OpenAI says it is also exploring new training methods to help models learn these values and understand why its principles matter, and plans to share results from that work in future updates. The company refined the Spec through research, deployment experience, public feedback, and expert discussions.