OpenAI reorganizes its Model Spec around five core values
Original title: An updated Model Spec for more capable and beneficial AI
The October 29 revision adds context, oversight, and agency guidance and retires much detailed style guidance.
OpenAI updated its public Model Spec on October 29, 2026, reorganizing behavior guidance around faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care. The revision explains existing principles more clearly, adjusts some behavioral defaults, and adds guidance on context, oversight, and agency. OpenAI says it is also exploring training methods so models apply those principles in new situations, and will share results later. Firm safety boundaries remain, while many other rules are defaults that users and developers can override.
Unconfirmed: OpenAI has not yet published results from the new training methods it says it is exploring.
- Five values are intent, harm minimization, honesty, agency, and care.
- OpenAI writes the Spec; developers and users direct work inside those boundaries.
- A new section treats instructions as the clearest signal of what people value.
- Training results and lessons are promised in future updates, not in this post.
From the original article
7 min read at openai.comOpenAI’s updated Model Spec outlines five core values to guide more capable AI toward sound judgment, human agency, and faithfulness to human intent.
openai.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.