OpenAI Updates Model Spec, Making Five Core Values the Guiding Principles for GPT Behavior

October 30 — OpenAI today released a new version of its "Model Spec," reorganizing the guidance system for GPT's behavior into five core values: being faithful to human intent, reducing harm, honesty and transparency, preserving human autonomy, and caring for the people affected. The new version adds guidance on contextual understanding, human oversight and user autonomy, requiring models not only to carry out the literal content of instructions but also to understand users' true goals and long-term interests, while avoiding steering their choices without the user's consent. OpenAI also adjusted instruction priority and default behavior rules, and removed most detailed writing style specifications, allowing models to adjust their expression according to user preferences, language and culture, and the specific context. In addition, the spec explicitly acknowledges that the models currently put into production do not yet fully meet the behavioral standards required by the new version, and that the company will continue to improve its models through training and feedback from real-world deployment.