OpenAI’s October 29 Model Spec update organizes intended behavior around human intent, avoiding harm, honesty and transparency, agency, and care. It emphasizes user control and oversight when models use context, memory or tools that affect the outside world. The revision places more weight on judgment across situations than on detailed stylistic prescriptions. It describes a behavioral target rather than certifying full compliance by current products. Evaluations should therefore distinguish the published principles from observed system behavior.