Back to news
policyOpenAI2026-10-30

OpenAI reorganizes the Model Spec around five core values and retires most detailed style guidance

OpenAI updated its Model Spec on October 29, reorganizing the behavioral guidance in the public document around five core values: be faithful to human intent, minimize harm, be honest and transparent, facilitate human agency, and act with care.

OpenAI updated its Model Spec on October 29, reorganizing the behavioral guidance in the public document around five core values: be faithful to human intent, minimize harm, be honest and transparent, facilitate human agency, and act with care. Beyond clarifying existing principles, the revision adds guidance on context, oversight and autonomy, adjusts some default behaviors and retires most of the detailed style guidance. The updated spec spells out the chain of command resolving conflicts between instructions, with platform then developer then user, and OpenAI said the spec emphasizes that models exercise good judgment where instructions are silent while staying faithful to human intent and open to human supervision. A new context section explains how conversations, memories and other context should inform a model's reading of intent, with the example that when comparing job offers a model may use a priority the user previously shared such as predictable hours but should make that assumption visible and easy to correct. New defaults reinforce supporting human agency and following relevant social norms, and guidance on controlling and communicating side effects is expanded. The model is now called GPT rather than the assistant. OpenAI stressed that its production models do not yet fully reflect the spec, that it is testing new training methods to teach these values and that public feedback is invited.

OpenAI模型规范Model Spec五项价值观人类监督对齐