What happened
On Oct 29, 2026, OpenAI published an update to its Model Spec, the public document that describes how it wants its models to behave. The guidance is now organized around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care.
OpenAI calls faithfulness to human intent the foundation. In its words, that means going beyond following instructions literally to understanding what OpenAI, developers and users intend, and acting wisely when that understanding is incomplete. The chain of command stays: OpenAI sets the boundaries, developers direct their apps within them, and users direct the work within those limits.
What's new and what's gone
A new subsection explains how conversations, memories and other context should shape the model's sense of what you want. OpenAI's own example: when helping you compare job offers, the model might use priorities you shared before, such as predictable hours, while making those assumptions visible and easy to correct.
There are new defaults on supporting human agency and following social norms, and expanded guidance on controlling and communicating side effects, which matters more as models take actions rather than just answer. Most of the detailed style guidance is retired, because OpenAI says style should adapt to what people prefer. One default stays firm: the model should approach questions objectively and shouldn't take a side just because it guesses that's what you want to hear. The Spec also now calls the model "GPT" instead of "the assistant".
What it doesn't promise
This is a description of behavior OpenAI is working toward, not a product release. OpenAI says plainly that its models do not yet fully reflect the updated Spec, and that it is exploring new training methods and will share results later. So don't expect ChatGPT to change overnight, and don't treat the Spec as a guarantee about any single answer.
OpenAI also says most of the Spec is made of defaults that users and developers can override, with firm limits only where needed to protect people.
What to do with it
If you use ChatGPT's memory, the new context guidance is a reason to check what it remembers about you and correct anything wrong. If you want a particular tone or point of view, say so directly; the Spec says the model shouldn't infer that from what it thinks you want to hear. OpenAI is asking for feedback on the Spec, linked from its post.