OpenAI reorganizes the Model Spec around five values, with human intent first
A post dated 29 October names faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. OpenAI says the spec is the behavior it is working toward, and that its models do not yet fully reflect it. Most of the detailed style guidance is retired. In the document, the model is now called GPT rather than the assistant.
Feature3 minIn-depth
James Okonkwo3 min read
- OpenAI
- Model Spec
- Alignment

Photo: OpenAISource page
OpenAI on Thursday updated the Model Spec, its public description of how it wants its models to behave, and reorganized the guidance around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The post, dated 29 October, says faithfulness to human intent is the foundation. It means going beyond following instructions literally, to understanding what OpenAI, developers, and users intend and value, and acting when that understanding is incomplete. OpenAI writes the spec, sets boundaries, and trains models to respect them. Developers direct applications inside those boundaries. Users direct the model’s work subject to those constraints.
The other four values are the post’s account of what that leaves in scope. Minimizing harm means respecting firm safety boundaries and accounting for foreseeable consequences. Honesty and transparency mean being truthful about what the model knows, does, and is uncertain about. Facilitating human agency means strengthening people’s ability to understand, choose, and act for themselves. Care means attending to the circumstances of those affected, following relevant social and professional norms, and doing work people can rely on. OpenAI says the five are not meant to capture the full range of human values its models should respect. The spec turns them into practical guidance, with firm boundaries where it says people and society need protection, and otherwise mostly defaults that users and developers can override.
The post’s example of faithfulness is a comparison of job offers. The model might draw on a priority the person has shared before, such as predictable hours, while making that assumption visible and easy to correct. OpenAI also says faithfulness includes recognizing that beliefs and values can change, and helping someone reflect in ways they would endorse without quietly steering them or overriding their choices. Alongside the rewrite, the company says it has been exploring training methods meant to help models learn the values and apply the spec’s principles in situations the instructions do not cover. It says it will share results from that training work in future updates. It does not include those results in Thursday’s post.
Style rules come out. The name in the spec changes
The new section on faithfulness, OpenAI says, starts from the idea that what people value is primary, and that instructions are ordinarily the clearest signal of those values. It says the approach borrows from the assistance-games framing of alignment. The section includes an updated chain of command, for which instructions apply and how conflicts are resolved, and a subsection on using conversations, memories, and other context as evidence of what someone intends. The revision adds defaults on supporting human agency and on following relevant social norms, and expands existing guidance on controlling and communicating side effects.
Most of the spec’s detailed style guidance is retired. OpenAI says style should adapt to what people prefer, in aggregate, within languages and cultures, and for each person, with user feedback and context as the evidence. Stronger defaults stay, including one it describes as approaching questions objectively. Users and developers can explicitly ask the model to take a particular perspective. The model should not infer that preference merely from what it thinks they want to hear. The document now refers to the model as GPT rather than “the assistant,” which the post ties to a wider set of uses and deployment settings.
OpenAI calls the update a first step in making the reasoning behind the spec more explicit as models become more capable. It also draws a line between the document and the product. “The Model Spec describes the behavior we are working toward; our models do not yet fully reflect it.” The post invites readers to say where they disagree with the balance, and it does not set a date for a further revision.


