What changed
On October 29, 2026, OpenAI announced an update to its Model Spec, which outlines how its AI models should behave. The revision reorganizes the guidance around five core values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. This update reflects OpenAI's ongoing work on value alignment, aiming to ensure models can understand and apply human values in unfamiliar situations while remaining faithful to human intent.
What this means in practice
The updated Model Spec provides clearer explanations of existing principles and adds guidance on context, oversight, and agency. It also adjusts some behavioral defaults and retires much of the detailed style guidance previously included. The new section on faithfulness to human intent emphasizes understanding what people value and acting wisely when that understanding is incomplete. OpenAI also introduced a new chain of command to explain which instructions apply and how conflicts are resolved. Additionally, the Spec now refers to the model as 'GPT' rather than 'the assistant,' reflecting the increasing breadth of its use cases and deployment settings.
Limitations and availability
The update is part of OpenAI's ongoing efforts to refine the Model Spec through research, user feedback, and discussions with experts. While the revised Spec provides practical guidance, it is not yet fully reflected in the models themselves. OpenAI acknowledges that the models do not yet fully embody the values outlined in the Spec and plans to continue improving both the Spec and the models as they learn from real-world use and feedback. The updated Model Spec is available for review, and OpenAI invites users to share their perspectives and feedback.
