OpenAI Unveils Latest Model Spec

OpenAI has released a major update to the Model Spec, which defines how AI models should behave. This update focuses on ensuring the model's customizability, transparency, and intellectual freedom without artificial restrictions.

The new framework clearly defines the priority of instructions from the platform (OpenAI), developers, and users through a Chain of Command. This allows users and developers to freely adjust the model's behavior within safe boundaries.

The key principles are as follows:

• Seek the truth together: Maintain objectivity, understand the user's goals, and provide critical feedback when necessary
• Do the best work: Adhere to fundamental capabilities such as factual accuracy, creativity, and programming ability
• Stay in bounds: Respect user autonomy while blocking harmful or potentially abusive requests
• Be approachable & Use appropriate style: Provide a friendly and empathetic conversational style along with clear formatting

In particular, this version explicitly emphasizes Intellectual freedom, allowing the free exchange of information even on controversial topics without forcing a specific intent. Additionally, it has been released under a CC0 license so that developers and researchers can freely use and modify it.