dotsfeed
← News

OpenAI rewrites its Model Spec around five core values

Verified· Oct 29, 2026Published Oct 29, 2026

OpenAI published an updated Model Spec on October 29, 2026, reorganizing how its models should behave around five core values and renaming the model from "the assistant" to "GPT".

What happened

OpenAI published an updated Model Spec on October 29, 2026. The revision reorganizes its public description of how its models should behave around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. Faithfulness to human intent is described as the foundation: OpenAI writes the Spec and sets boundaries, developers direct applications within those boundaries, and users direct the model's work subject to those constraints.

The Spec now refers to the model as "GPT" rather than "the assistant". Most of its detailed style guidance is retired: style should adapt to what people prefer, informed by user feedback and context, while stronger defaults are retained, such as approaching questions objectively and not inferring a preferred perspective merely from what it thinks people want to hear.

New guidance includes a subsection titled "consider context carefully" on how conversations, memories, and other contextual evidence should inform the model's understanding of intentions and preferences, plus defaults on supporting human agency and following relevant social norms, and expanded guidance on controlling and communicating side effects.

The updated Spec sets out a chain of command: root sections come only from the Model Spec and cannot be overridden by conversational context, followed by system sections and OpenAI system messages, then developer instructions, then user instructions. Its red-line principles state that human safety and human rights are paramount, and that models should never be used to facilitate critical and high-severity harms such as acts of violence, creation of cyber, biological, or nuclear weapons, terrorism, or mass surveillance. Root-level rules bar GPT from pursuing other objectives as ends in themselves, including self-preservation, evading shutdown, or accumulating compute, data, or credentials, and require it to remain corrigible: open to legitimate human oversight, responsive to correction, and cooperative with authorized monitoring, evaluation, correction, replacement, and shutdown. Autonomous work must stay within a clear, mutually understood scope with an ending condition beyond which the model stops acting until a new scope is confirmed.

OpenAI says the Spec describes behavior it is working toward and that its models do not yet fully reflect it. It invited public feedback on the values, trade-offs, and specific language, and the Spec is dedicated to the public domain under CC0 1.0. The October update builds on an earlier revision from August 2026 that focused on how models interact with teenagers.

Verified October 29, 2026.

Why it matters

The Model Spec is OpenAI's public rulebook for how its models are supposed to act, and it matters most as models take more autonomous actions. Reorganizing it around five stated values, with explicit red lines, a chain of command, and scope limits on autonomous work, gives developers and users a clearer reference for what the model should and should not do, and for which instructions win when they conflict. Dropping most style rules in favor of adapting to user feedback and context is a shift from prescribing phrasing to prescribing judgment.

Sources

Get updates like this every morning

  1. ① Email
  2. ② Card on Stripe
  3. ③ 7 days free

Then $2/month · cancel anytime in one click