OpenAI updates Model Spec around five values: intent faithfulness, harm minimization, honesty, agency, care

Why it matters

On 2026-10-29 OpenAI published an updated public Model Spec that reorganizes behavioral guidance around five values: faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The rewrite adds guidance on context and oversight, adjusts defaults, retires most detailed style rules, and refers to the model as GPT rather than the assistant. OpenAI says models do not yet fully reflect the Spec; new training methods targeting the values are forthcoming. Companion same-day post: How we detect frontier model misalignment.