OpenAI Model Spec Update: The 5 Values Now Guiding GPT

OpenAI has published a significant revision to its Model Spec, the public document that describes how it wants its models to behave. Announced on October 29, 2026, the update reorganizes the whole Spec around five core values instead of a long list of rules, adds new guidance on context and human agency, and retires most of the detailed style instructions that used to shape how ChatGPT writes.

• Five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care
• New guidance: how to use context such as memories, plus defaults on supporting human agency and following social norms
• Retired: most of the Spec's detailed style guidance
• Naming change: the Spec now calls the model "GPT" rather than "the assistant"

What Is the OpenAI Model Spec?

The Model Spec is OpenAI's written description of intended model behavior: what its models should do, what they should refuse, and how they should resolve conflicts between instructions from OpenAI, developers and users. It is public, which means anyone can read the rules a model is trained toward and push back on them. OpenAI is clear that the Spec "describes the behavior we are working toward" and that current models do not yet fully reflect it.

The Five Core Values, Explained

According to OpenAI's announcement, the five values are not new commitments so much as a clearer organizing frame for principles already scattered through the document:

• Faithfulness to human intent: going beyond literal instructions to understand what people actually want, and acting sensibly when that understanding is incomplete. OpenAI calls this the foundation of the approach
• Minimizing harm: respecting firm safety boundaries and thinking about the foreseeable consequences of the model's actions
• Honesty and transparency: being truthful about what the model knows, does and is unsure about
• Human agency: strengthening people's ability to understand, choose and act for themselves
• Care: paying attention to the circumstances of the people affected and producing work people can rely on

OpenAI gives a practical example of faithfulness: when helping someone compare job offers, the model might draw on priorities the person shared earlier, such as predictable hours, while making those assumptions visible and easy to correct.

What Changed in This Model Spec Update?

The new section on faithfulness to human intent folds in the updated "chain of command," which explains whose instructions apply and how conflicts get resolved. A new subsection on considering context explains how conversations, memories and other signals should inform the model's understanding of what someone wants.

The more visible change for everyday users is the retirement of most style guidance. OpenAI says style should adapt to what people prefer, in aggregate, across languages and cultures, and per person. Some firm defaults stay, including approaching questions objectively. Users and developers can ask for a particular perspective, but the model should not assume one simply because it thinks that is what someone wants to hear.

Why Value-Based AI Guidance Matters

Rules work well for situations their authors predicted. As models take on longer, more agentic tasks, they keep running into situations nobody predicted. Grounding behavior in a few explained values is an attempt to help models generalize sensibly. OpenAI says it is also exploring training methods that teach models the reasoning behind the Spec, with results to be shared later.

For developers building on OpenAI's API, such as the newer GPT-6.1 Terra tier, the Spec is worth reading in full: it defines what your app can override and what it cannot. OpenAI is inviting public feedback on the values and trade-offs. More on model behavior and alignment research lives in our AI section.

Sources: OpenAI: An updated Model Spec for more capable and beneficial AI — October 29, 2026; OpenAI: Our approach to the Model Spec — background.