OpenAI updates its Model Spec around five core values for AI behavior

The five values

OpenAI published an updated Model Spec on 29 October 2026, reorganizing the framework around five core values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The Spec is a public framework — dedicated to the public domain under CC0 1.0 — that describes the behavior OpenAI wants its GPT models to exhibit.Faithfulness to human intent is positioned as the foundation. The Spec says models must go beyond following instructions literally to deeply understanding what OpenAI, developers and users intend and value. That means recognizing when instructions conflict with underlying values, making assumptions visible, and acting wisely when understanding is incomplete. The other four values are presented as making the implications of faithfulness explicit: minimize harm, be truthful, respect human freedom to choose, and attend to the circumstances of affected people.

Chain of command and context

The Spec introduces an updated chain of command that resolves conflicts between instructions: OpenAI writes the Spec and sets boundaries; developers direct applications within those boundaries; users direct the model within those constraints. The model should be faithful to each level while recognizing that higher-level instructions override lower-level ones.

A new subsection on context explains how conversations, memories and other contextual evidence should inform the model's understanding of intentions and preferences. The Spec adds defaults for following relevant social norms and supporting human agency, and expands guidance on controlling and communicating side effects of the model's work. It also explicitly retires most detailed style guidance — letting user feedback shape style — while retaining some defaults such as approaching questions objectively.

What it means

The Model Spec update is significant because it moves OpenAI's thinking on AI behavior from a narrower instruction-following framework to a value-alignment framework expressed in explicit principles. As models get smarter and take more autonomous actions — running code, calling tools, managing workflows — the question of what principles govern their behavior becomes operationally important, not just philosophical.

The shift from 'the assistant' to 'GPT' in the Spec language reflects the broadening deployment context: the same model now powers API calls, ChatGPT conversations, agentic tools and autonomous coding agents. OpenAI is acknowledging that a single behavior framework needs to apply across all those roles. The retraining effort implied — 'we are continually refining and updating our systems to bring them into closer alignment with these guidelines' — signals an active investment in value alignment as a product feature, not just a research topic.

Caveats

The updated Spec describes the behavior OpenAI is working toward, not behavior the current models already exhibit. OpenAI explicitly states production models do not yet fully reflect it. The Spec also intentionally leaves many details of style, personality and interaction unspecified.

The effectiveness of the Spec depends on how well training methods can translate the five values into consistent model behavior, and on how OpenAI handles edge cases where the values conflict. The Spec itself acknowledges these trade-offs and delegates resolution to the chain of command.

The training approach

Alongside the Spec update, OpenAI says it has been exploring new training methods to help models learn these core values — to do 'the right things for the right reasons.' The aim is to build an understanding of why the principles matter and how to apply them in new situations, not just to follow a fixed rule set. This suggests a shift from rule-based alignment to value-based alignment in training methodology.

The Spec describes the behavior OpenAI is working toward, and the company is clear that current production models do not fully reflect it. The retraining work is ongoing and the Spec will be continuously updated based on feedback and lessons from serving users worldwide.

What was retired

The update retires most of the Spec's detailed style guidance — the precise instructions about tone, format and phrasing that earlier versions contained. The reasoning is that style should adapt to what people prefer, in aggregate, within languages and cultures, and for each person. User feedback and context provide evidence about those preferences.

The Spec retains stronger defaults such as approaching questions objectively, and users and developers can explicitly ask the model to take a particular perspective. But the model should not infer that preference merely from what it thinks they want to hear — a guard against sycophancy that remains in the framework.

Public domain and feedback

The Spec is dedicated to the public domain under the Creative Commons CC0 1.0 deed, a choice intended to encourage wide use and collaboration. OpenAI explicitly invites feedback on which parts of the Spec are compelling, where the balance it strikes is wrong, and what it overlooks or leaves unclear.

The Spec is complemented by the company's usage policies (which define expectations for how people should use the API and ChatGPT) and its safety protocols (testing, monitoring and mitigation). Together these three layers define the behavior and accountability framework for OpenAI's deployed models.