OpenAI Open Source Teen Safety Prompts for AI Developers

OpenAI announced the release of OpenAI teen safety prompts, open source tools designed to help developers build safer AI applications for teenagers. The initiative responds to rising scrutiny over AI’s impact on youth mental health and safety. The prompts are practical safety policies that developers can integrate into their applications. They are built to work with OpenAI’s open-weight safety model, gpt-oss-safeguard, while staying compatible with other AI models. OpenAI teen safety prompts target key risk areas, including graphic violence and sexual content, harmful body ideals linked to eating disorders, dangerous challenges circulating online, and romantic or violent role-play scenarios. The framework also covers age-restricted goods and services. OpenAI said it developed the prompts with Common Sense Media and everyone.ai to ensure the rules reflect current research and real-world implementation needs. Common Sense Media’s Robbie Torney described the approach as creating a “safety floor” across the ecosystem, with the advantage that open source tools can be adapted and improved over time. The company also highlighted why safety implementation is hard for developers: teams often struggle to translate broad safety goals into clear, operational rules, leading to gaps or overly broad filtering. OpenAI teen safety prompts are positioned as tested, scoped guidance that can reduce development time. This release builds on OpenAI’s earlier youth protections, including parental controls and age prediction features, and its Model Spec updates for under-18 user interactions. OpenAI cautioned that the prompts are not a complete solution, but one component within a broader safety ecosystem amid increasing regulatory pressure such as the EU AI Act.