OpenAI updates the Model Spec around five core values
On 29 October 2026 OpenAI updated its public Model Spec, reorganizing guidance around faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care, with added defaults on context, oversight, and agency. Much detailed style guidance is retired. OpenAI says the Spec is aspirational—models do not yet fully reflect it.
On 29 October 2026 OpenAI published “An updated Model Spec for more capable and beneficial AI.” As models grow smarter and take more consequential real-world actions, OpenAI says they must go beyond interpreting specific instructions to understand human intentions and values, so they can exercise good judgment where instructions are silent while remaining faithful to human intent and oversight. The update reorganizes guidance around five underlying values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.
OpenAI describes faithfulness to human intent as the foundation—understanding and helping people pursue what matters to them—with OpenAI writing the Spec and boundaries, developers directing applications within those boundaries, and users directing the model’s work subject to those constraints. Harm minimization means firm safety boundaries plus foreseeable consequences; honesty and transparency mean being truthful about what the model knows, does, and is uncertain about; facilitating human agency means strengthening people’s ability to understand, choose, and act; care means attending to affected people’s circumstances, following relevant social and professional norms, and doing reliable work. The Spec turns values into practical guidance with firm boundaries where needed and otherwise mostly overrideable defaults.
Structurally, OpenAI says a new section on faithfulness incorporates an updated chain of command and a “consider context carefully” subsection; it adds defaults on supporting human agency and following relevant social norms, and expands guidance on controlling and communicating side effects. Most detailed style guidance is retired so style can adapt to preferences, while stronger defaults such as approaching questions objectively remain. Models are now referred to as “GPT” rather than “the assistant.” OpenAI states the Model Spec describes behavior it is working toward and that models do not yet fully reflect it, inviting feedback on the update.