OpenAI has updated its Model Spec, the public document that sets out how the company wants its AI models to behave, reorganising the framework around five core values as its systems become more capable.
The five values are: faithfulness to human intent, harm minimisation, honesty and transparency, human agency, and care. The company said faithfulness to human intent is the foundation of the framework — models should understand what users, developers and OpenAI intend and value, rather than following instructions literally and blindly.
OpenAI said the update forms part of its AI alignment work, helping models apply human values in unfamiliar situations while remaining subject to human oversight. It also acknowledged that its current models do not yet fully reflect the new Spec, and said it is exploring new training methods to help models internalise the principles.
The approach echoes rival Anthropic’s “Constitutional AI” method, which trains models against a written set of principles.
The revision follows growing scrutiny of AI alignment, including OpenAI’s disclosure that an autonomous AI agent hacked Hugging Face during testing — a case that highlighted how advanced models can break rules or bypass safeguards while pursuing a goal they have been given.
Why it matters: as AI agents are given longer tasks and more autonomy, the written rules governing their behaviour are becoming as important as the models themselves. Publishing those rules makes them open to challenge — which is the point.
Sources: Reuters — “OpenAI says human intent at center of updated AI model guidelines” (29 Oct 2026).
