Teen Safety Policy Pack Available Now
OpenAI has released a set of prompt-based safety policies designed to help developers create safer AI experiences for teenagers. These policies are built to work with OpenAI's open-weight safety model, gpt-oss-safeguard-20b, and translate high-level safety requirements into practical, operational rules that can be deployed in real-world systems.
The initial policy release covers six critical areas:
- Graphic violent content
- Graphic sexual content
- Harmful body ideals and behaviors
- Dangerous activities and challenges
- Romantic or violent roleplay
- Age-restricted goods and services
Why This Matters for Developers
A major challenge in AI safety is translating abstract safety goals into precise, enforceable rules. Even experienced teams struggle to define policies that accurately capture teen-specific risks while avoiding over-blocking legitimate content. By releasing these policies as structured prompts, OpenAI enables developers to more easily integrate consistent safety standards into existing workflows and adapt them to their specific use cases.
These policies can be used for both real-time content filtering during active interactions and offline analysis of user-generated content, making them flexible for different deployment scenarios.
Developed With External Expertise
OpenAI collaborated with domain experts including Common Sense Media and everyone.ai to inform the development of these policies. Their input helped shape the scope of coverage, strengthen the prompt structure, and refine edge cases. This reflects a broader commitment to building safety infrastructure that benefits the entire ecosystem.
Getting Started
The policies are released as open source through the ROOST Model Community on GitHub. Developers are encouraged to adapt and extend these policies based on their specific needs and combine them with complementary safeguards like thoughtful product design, user controls, monitoring systems, and age-appropriate response mechanisms. OpenAI emphasizes that these policies are a starting point, not a comprehensive solution—each application should evaluate its unique risks and audience contexts.