Independent newsroom

Friday, October 30, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

Labs · Friday, October 30, 2026

OpenAI rewrites its model rulebook around five values

The File desk · Oct. 30, 2026, 6:37 a.m. ET

Confirmed1 primary sources

Status

Confirmed. OpenAI published the Model Spec update on October 29, 2026. The five values, the retired style guidance, and the statement that models do not yet fully match the spec are the company's account of its own document.
  • Confirmed
    The post lists five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.
  • Confirmed
    OpenAI says the spec describes behavior it is working toward and that its models do not yet fully reflect it.
OpenAI on Thursday updated the Model Spec, the public document that describes how it wants its models to behave. The revision organizes that guidance under five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The company said its models do not yet fully reflect the spec.1
The company said models need to understand what people intend when a written instruction does not cover the case, and still stay faithful to human intent and oversight. OpenAI writes the spec and sets boundaries. Developers direct applications inside those boundaries. Users direct the work, subject to the same limits.1
The revision adds guidance on context, oversight, and agency. It retires most of the detailed style guidance. Style, the post says, should follow what people prefer. One stronger default stays: approach questions objectively. The spec now calls the model “GPT” rather than “the assistant.”1
Why it matters (Estimate). Estimate. A person using a chatbot on Thursday cannot treat this document as a description of what the bot already does. OpenAI says the spec is the behavior it is working toward, and that current models do not fully match it. The five values are the company's list, not a statute. No deadline is printed for when the models will catch up.
“The Model Spec describes the behavior we are working toward; our models do not yet fully reflect it.” The post invites readers to comment. It does not give a date for when the models will match the spec.1
How we know

Notes

  1. 1.OpenAI, “An updated Model Spec for more capable and beneficial AI,” October 29, 2026. https://openai.com/index/updated-model-spec “Today we’re updating the Model Spec, our public description of how we want our models to behave. The update reorganizes its guidance around five underlying values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.” The post says the revision adds guidance on context, oversight, and agency, adjusts some behavioral defaults, and retires much of the detailed style guidance. “Style should adapt to what people prefer.” “The Spec retains stronger defaults, such as approaching questions objectively.” “We now refer to the model as ‘GPT’ rather than ‘the assistant’ in the Spec.” “The Model Spec describes the behavior we are working toward; our models do not yet fully reflect it.”

What is still unknown or disputed

  • The full text of the spec, as distinct from this announcement. The post links the spec. This story uses the announcement.
  • When any particular product will follow the new wording. No date is given.
  • The Luna addendum posted the same day. It is a separate note about one model's answers, not this rulebook.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    An updated Model Spec for more capable and beneficial AI

    OpenAI · October 29, 2026

    OpenAI reorganized the Model Spec around five values and said its models do not yet fully reflect the document.

    https://openai.com/index/updated-model-spec