OpenAI has introduced a specialized suite of safety policies designed specifically for developers building AI experiences for teenagers. Rather than relying on rigid hard-coded rules, these policies are structured as natural language prompts for the company’s open-weight safety model, gpt-oss-safeguard. By providing these templates, the initiative aims to simplify the process of creating classifiers—automated systems that categorize and filter content—ensuring that AI interactions remain age-appropriate and secure.

The initiative focuses on six critical areas of concern for younger users: graphic violence, harmful body ideals, dangerous behavioral challenges, romantic roleplay, and access to age-restricted goods. These prompts allow developers to implement real-time filtering or perform offline audits of user-generated content with greater precision than general-purpose safety filters. This reflects a growing industry recognition that teenagers require distinct digital protections and more nuanced moderation compared to adult users.

To foster a transparent and collaborative environment, the policies have been released through the ROOST Model Community. This open-source approach encourages developers to adapt the frameworks to local languages or specific cultural contexts. By collaborating with organizations like Common Sense Media, OpenAI is attempting to bridge the gap between high-level safety principles and the technical tools needed to enforce them in production environments.

OpenAI has introduced a specialized suite of safety policies designed specifically for developers building AI experiences for teenagers. Rather than relying on rigid hard-coded rules, these policies are structured as natural language prompts for the company’s open-weight safety model, gpt-oss-safeguard. By providing these templates, the initiative aims to simplify the process of creating classifiers—automated systems that categorize and filter content—ensuring that AI interactions remain age-appropriate and secure.

The initiative focuses on six critical areas of concern for younger users: graphic violence, harmful body ideals, dangerous behavioral challenges, romantic roleplay, and access to age-restricted goods. These prompts allow developers to implement real-time filtering or perform offline audits of user-generated content with greater precision than general-purpose safety filters. This reflects a growing industry recognition that teenagers require distinct digital protections and more nuanced moderation compared to adult users.

To foster a transparent and collaborative environment, the policies have been released through the ROOST Model Community. This open-source approach encourages developers to adapt the frameworks to local languages or specific cultural contexts. By collaborating with organizations like Common Sense Media, OpenAI is attempting to bridge the gap between high-level safety principles and the technical tools needed to enforce them in production environments.