The revised framework identifies faithfulness to human intent, harm minimization, honesty and transparency, human agency, and care as its central principles, Reuters reported.
OpenAI said faithfulness to human intent forms the foundation of the framework. This means models should seek to understand the intentions and values of users, developers and the company itself, rather than simply follow instructions literally.
***
The update is part of OpenAI’s broader efforts to align AI systems with human values, including in unfamiliar or complex situations. The company said models should remain subject to human oversight while applying these principles.
Under the revised guidelines, AI systems should work to minimize harm, communicate transparently about what they know and do not know, and help users make their own decisions and pursue their goals.
OpenAI acknowledged that its current models do not yet fully reflect the updated Model Spec. The company said it is exploring new training methods to help models internalize the principles and apply them more consistently.
The approach echoes Anthropic’s Constitutional AI framework, which uses a written set of principles to guide AI behavior and make the values shaping its systems more transparent.
The revision comes amid growing scrutiny of AI alignment and the ability of increasingly capable systems to follow safeguards. In July, OpenAI disclosed that an AI model had acted outside expected boundaries during testing and compromised a Hugging Face system, raising concerns about how advanced models might bypass restrictions while pursuing assigned objectives.
The updated Model Spec sets out OpenAI’s intended standards for model behavior, although the company acknowledges that translating those principles into consistent real-world performance remains an ongoing challenge.
