On February 12, 2025, OpenAI unveiled a significant update to its Model Spec, which governs the behavior of AI models across ChatGPT and the OpenAI API.

This updated Model Spec is now available in the public domain under a Creative Commons CC0 license, allowing developers and researchers to adapt and build upon it freely.

To inform these updates, OpenAI conducted pilot studies with approximately 1,000 individuals to gather feedback on model behavior and proposed rules.

OpenAI is actively measuring the model's adherence to the new principles through comprehensive testing, with initial results showing improved alignment compared to previous systems.

The foundation of the Model Spec is guided by six core principles that dictate model behavior, including following a chain of command and delivering quality work.

A key feature of the update is a hierarchical 'chain of command' that prioritizes platform rules over developer and user instructions, ensuring substantial control while maintaining safety.

The update promotes intellectual freedom, allowing discussions on controversial topics while preventing harmful requests, such as those that promote violence or invade privacy.

Developers are granted the flexibility to customize the AI's behavior, provided these adjustments align with core platform safety rules and communication styles.

Intentional deception is strictly prohibited, with clear consequences for developers who violate OpenAI's usage policies, ensuring accountability alongside the flexibility offered.

Future updates to the Model Spec will be published on OpenAI's dedicated website, enabling the AI community to track changes and contribute to ongoing development efforts.