← Back
OpenAI
OpenAI releases teen safety policies for gpt-oss-safeguard model
OpenAI APIOpenAI · releasefeaturesecuritysdkopen-source · openai.com ↗

Prompt-Based Safety Policies for Teen Protection

OpenAI has released a comprehensive set of teen safety policies as open-source resources available through GitHub. These policies are specifically designed to work with gpt-oss-safeguard, OpenAI's open-weight safety model, enabling developers to implement consistent age-appropriate protections across their applications.

Six Coverage Areas

The initial policy release addresses the most common teen-specific risks:

  • Graphic violent content and dangerous activities
  • Graphic sexual content and romantic/violent roleplay
  • Harmful body ideals and behaviors
  • Dangerous activities and challenges
  • Age-restricted goods and services

Each policy is structured as a prompt that can be integrated directly into existing safety workflows, supporting both real-time content filtering and offline analysis of user-generated content.

Expert-Informed Development

OpenAI developed these policies in collaboration with Common Sense Media and everyone.ai, organizations with deep expertise in youth protection and AI safety. This external input helped shape the scope, strengthen the prompt structures, and refine edge cases for practical application.

Developer Integration and Iteration

By structuring policies as prompts, developers can more easily integrate them into existing systems, adapt them to specific use cases, and iterate over time. OpenAI emphasizes that these policies serve as a starting point—a meaningful safety baseline—rather than a comprehensive final solution. Developers should adapt and extend these policies based on their specific contexts and combine them with additional safeguards including product design decisions, user controls, and monitoring systems.

The policies are released through the ROOST Model Community as open source, encouraging broad ecosystem collaboration and continuous improvement.