OpenAI releases new version of Model Spec, enhancing customizability and transparency

According to an announcement by OpenAI, the company has released a significant update to Model Spec, a document that defines the behavioral standards of AI models. The new version emphasizes customizability, transparency, and freedom of thought, ensuring users can freely explore, discuss and create while maintaining necessary safety measures. The updates include:

• Model instruction priority: Clarifying the instruction priorities of platforms, developers and users to enhance customization capabilities.

• Information authenticity: Models avoid biased guidance and encourage objective discussions to help users make independent decisions.

• Safety boundaries: Setting standards for refusing high-risk requests such as harmful behavior or privacy violations.

• Knowledge sharing: This specification has been open-sourced to the public domain so developers are free to use, modify and build upon it.

OpenAI stated that this specification will continue to evolve and welcomes feedback from the community.