OpenAI centers updated Model Spec on human intent

OpenAI reorganized its public Model Spec around five values, putting faithfulness to human intent at the center while acknowledging that its current models do not yet fully follow the guidance. The company announced the update on October 29, 2026. The five values are faithfulness to human intent, minimizing harm, honesty and transparency, human agency, and care. The document sets out how OpenAI wants its models to behave and asks them to consider the goals and values behind requests, rather than following wording literally. OpenAI said the framework is intended to help models apply its principles in unfamiliar situations while remaining subject to human oversight. The guidance calls for limiting foreseeable harm, being transparent about what a model knows and does not know, supporting people’s choices, and considering others affected by model outputs. The revised Spec distinguishes firm safety boundaries from default behaviors that users or developers can change. OpenAI also said it added guidance on context, oversight and agency, adjusted some defaults, and removed much of the earlier detailed advice on style. The document now refers to the model as GPT rather than “the...