OpenAI reorganizes its Model Spec around five values
OpenAI's public rulebook for how its models should behave now centers on five values. Here is what that means for the AI tools your business uses, and three things to do this week.
What OpenAI changed
On Oct 29, 2026, OpenAI published An updated Model Spec for more capable and beneficial AI. In OpenAI's words, the Model Spec is "our public description of how we want our models to behave." You can think of it as the rulebook OpenAI writes for its models and trains them toward.
The update does three main things. It reorganizes the guidance around five underlying values. It adds guidance on context, oversight and agency. It also adjusts some behavioral defaults and retires much of the detailed style guidance. One smaller change: the Spec now calls the model "GPT" rather than "the assistant."
OpenAI's reasoning is easy to follow. As models get smarter and take more consequential actions, following the literal instruction is not enough. They also need to understand what people intend, so they can use good judgment in situations their instructions don't explicitly address, while staying faithful to human intent and oversight.
The five values
Here are the five values, each with OpenAI's own one-line meaning.
- 01
1. Faithfulness to human intent.
Understanding and helping people pursue what matters to them. OpenAI calls this the foundation of the approach.
- 02
2. Minimizing harm.
Respecting firm safety boundaries and accounting for the foreseeable consequences of the model's actions.
- 03
3. Honesty and transparency.
Being truthful about what the model knows, does, and is uncertain about so people can understand and evaluate its reasoning and work.
- 04
4. Facilitating human agency.
Strengthening people's ability to understand, choose, and act for themselves.
- 05
5. Care.
Attending to the circumstances of those affected, following relevant social and professional norms, and doing work people can rely on.
Value definitions are OpenAI's, from An updated Model Spec for more capable and beneficial AI (Oct 29, 2026).
OpenAI says these five are not meant to cover the full range of human values. They organize the Spec and guide how the models should handle some of the most important trade-offs they face.
Who gets to set the rules
The update includes a revised chain of command, which explains which instructions apply and how conflicts are resolved. In OpenAI's words: "OpenAI writes the Spec, sets boundaries, and trains models to respect them; developers direct applications within those boundaries; and users direct the model’s work subject to those constraints."
Much of the Spec can be changed. Outside the firm boundaries that protect people and society, it "mostly takes the form of defaults that users and developers can override." That matters when you use a tool another company built on top of OpenAI's models: the builder may have changed some defaults before you ever type a word.
What else is new
A new subsection called "consider context carefully" explains how conversations, memories and other contextual evidence should inform the model's understanding of people's intentions and preferences. OpenAI gives an example: when helping someone compare job offers, the model might draw on priorities they shared before, such as predictable hours, while making those assumptions visible and easy to correct.
The Spec also adds defaults on supporting human agency and following relevant social norms, and expands its guidance on controlling and communicating side effects. As OpenAI describes them, these cover things like considering the effects of the work on others and producing work people can rely on, while respecting the user's goals and instructions.
Most of the detailed style guidance is gone. OpenAI says style should adapt to what people prefer, with feedback and context as evidence. Some stronger defaults stay, such as approaching questions objectively. You can ask the model to take a particular perspective, but it should not assume you want one just because it thinks that is what you want to hear.
Why it matters to a small business
Most instructions people give an AI tool are short. "Draft a reply to this customer." "Clean up this price list." "Book the follow-up." Each one leaves gaps, and the model fills those gaps with its own read of what you meant.
This update is OpenAI's description of how that gap-filling should work. The model is meant to use context, including past conversations and memories, to infer your intent. It is meant to be honest about what it knows and does not know. It is also meant to consider how its work affects people other than you, such as your customers.
You can put that to use. If the model reads context to understand you, the context you give it matters. If it is supposed to make its assumptions visible, you can ask to see them. And since defaults can be changed, it is worth knowing who changed them in the tools you pay for.
Three things to do this week
These are our suggestions for an owner or manager, based on what the Spec describes. They are not OpenAI's instructions.
- 01
1. Write your intent down first.
Before a bigger AI task, write a few lines: what you want, who it affects, and what it should not touch. The Spec says the model should use context to understand intentions, so give it good context to work from.
- 02
2. Ask for assumptions and side effects before it acts.
For anything that goes to a customer or touches money, ask the AI to list the assumptions it is making and what else its work could change. Correct anything that is wrong, then let it go ahead.
- 03
3. Check the defaults in the tools you buy.
If an app is built on OpenAI's models, its developer can override many defaults within OpenAI's boundaries. Ask the vendor what they have set, and test the tool on a real task before you rely on it.
In all three cases, keep a person on decisions that affect customers or money.
If you are trying tools that take actions on a computer for you, our earlier piece on GPT-6.1 Terra and computer use covers that side.
Read with care
The Model Spec describes how OpenAI wants its models to behave. It does not promise how they behave today, and OpenAI says so directly: "The Model Spec describes the behavior we are working toward; our models do not yet fully reflect it."
Two more points. First, the developer who built an app can override the defaults, and so can users within limits, which means the same model may act differently in different tools. Second, OpenAI says it is exploring new training methods to help models learn these values and will share results in future updates. Until then, treat the five values as goals. This post is not legal advice.
Our take
For a small business, the most useful part of this update is a simple reminder: your instructions will always run out somewhere, and the model will fill the gap. Tell it what you want and what matters, ask it to show its assumptions, and find out what your tools have set. Anything that reaches a customer or your bank account still needs a person's final call.
Sources
An updated Model Spec for more capable and beneficial AI (OpenAI, Oct 29, 2026; verified live Oct 30, 2026)
OpenAI Model Spec, Oct 29, 2026 version (OpenAI, Oct 29, 2026; linked from the announcement. All facts in this post come from the announcement text.)
Want this working in your business?
Tell us what you are trying to fix and we will point you at the smallest useful next step.
Talk to us