OpenAI reorganizes its Model Spec around five core values

In short

OpenAI says it has updated its Model Spec, its public description of how it wants its models to behave. The new version is organised around five values. This is a change to written guidance, not a report of a test or incident, and OpenAI says its models do not yet fully follow the guide.

What happened

OpenAI announced the update on October 29, 2026, in the original report. The company says that as its models get smarter and take more consequential actions, they need to do more than interpret specific instructions. They also need to understand human intentions and values, so they can use good judgment when their instructions do not cover a situation.

The guidance is now arranged around these five values:

• Faithfulness to human intent: understanding what OpenAI, developers and users intend and value, rather than only following instructions literally.
• Harm minimization: respecting firm safety boundaries and accounting for foreseeable consequences of the model's actions.
• Honesty and transparency: being truthful about what the model knows, does and is uncertain about.
• Human agency: strengthening people's ability to understand, choose and act for themselves.
• Care: attending to the circumstances of those affected, following relevant social and professional norms, and doing reliable work.

OpenAI says these values reflect commitments already found across the Spec. It also says they are not meant to capture every human value its models should respect.

The update also adds and changes some guidance. A new section on faithfulness includes an updated "chain of command", which explains which instructions apply and how conflicts are resolved. A new subsection explains how conversations, memories and other context should inform the model's understanding of what people want. OpenAI has also added defaults, which are behaviours users and developers can usually override, on supporting human agency and following social norms. It expanded its guidance on controlling and communicating side effects.

OpenAI is also retiring most of the Spec's detailed style guidance. It says style should adapt to what people prefer. Some stronger defaults remain, such as approaching questions objectively. People can ask the model to take a particular perspective, but OpenAI says the model should not assume that preference merely from what it thinks they want to hear. The Spec now also calls the model "GPT" rather than "the assistant".

The report does not describe a training run, test, simulation or real-world incident. It does not say that any person or organisation was affected by a specific event.

What was done about it

OpenAI says it is exploring new training methods to help its models learn these values and understand why the Spec's principles matter. It says it will share results and lessons in future updates. It is also asking the public for feedback on the values, trade-offs and wording, including what the Spec overlooks or leaves unclear.

Why it matters

OpenAI says the Spec describes the behaviour it is working toward, and that its models do not yet fully reflect it. The company calls this update a first step in making the values and reasoning behind the Spec more explicit as models become more capable. It says it will keep improving both the Spec and its models as it learns from research, real-world use and feedback.