OpenAI updated its public Model Spec on Oct. 29, 2026, reorganizing its description of how it wants models to behave around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The company says models taking more consequential actions need to understand human intentions and values, not just follow specific instructions.
OpenAI calls faithfulness to human intent the foundation of the approach. In its account, OpenAI sets boundaries, developers direct applications within them, and users direct a model’s work subject to those constraints. The other values address foreseeable harm, honesty about what a model knows and does, people’s ability to make their own choices, and attention to those affected by the work. OpenAI says the five values are not meant to encompass every human value its models should respect.
What changed in the guidance
The revision includes an updated chain of command for determining which instructions apply and resolving conflicts. A new subsection addresses how conversations, memories and other context should inform a model’s understanding of a person’s intentions and preferences. OpenAI also added defaults on supporting human agency and following relevant social norms, and expanded guidance on controlling and communicating side effects.
OpenAI retired much of the Spec’s detailed style guidance, saying style should adapt to people’s preferences. It says stronger defaults remain, including approaching questions objectively. The Spec now calls the model “GPT” rather than “the assistant.”
The update describes intended behavior, not a result OpenAI claims to have achieved. OpenAI says its models do not yet fully reflect the Spec. It has been exploring new training methods to help models learn the values and says it will share results and lessons in future updates. The company also invites feedback on the Spec’s values, trade-offs and wording.
Comments