OpenAI is releasing prompt-based safety policies to help developers build age-appropriate protections for teens, including for use with gpt-oss-safeguard and other reasoning models.

The policies translate broad safety goals into clearer, more consistent rules developers can apply in real systems—an effort to reduce protection gaps, uneven enforcement and overly broad filtering.

The initial policies address graphic violence and sexual content, harmful body ideals and behaviors, dangerous activities and challenges, romantic or violent roleplay, and age-restricted goods and services; they can be used to filter content in real time or review user-generated content offline.

Common Sense Media and everyone.ai contributed feedback that helped shape the topics, prompt structure and edge cases.

OpenAI describes the policies as a starting point, not a complete definition or guarantee of teen safety, and encourages developers to tailor them to their products and combine them with user controls, monitoring, thoughtful product design and age-appropriate responses.

The policies are being released as open source through the ROOST Model Community, where developers and organizations can contribute, adapt, translate and extend them.