OpenAI is releasing a set of open-source safety policies aimed at helping developers build age-appropriate protections into AI applications used by teenagers, making the tools freely available for anyone building on its models.
The policies are formatted as prompts that developers can run with OpenAI's gpt-oss-safeguard model — a safety-focused open-weight model — to generate consistent content classifiers without having to build such systems from scratch.
The initial release targets six categories of risk relevant to younger users: graphic violent content, graphic sexual content, harmful body ideals and behaviours, dangerous activities and challenges, romantic or violent roleplay, and age-restricted goods and services. OpenAI says the prompts can be used both to filter content in real time and to analyse material offline.
The move extends a broader push by OpenAI on under-18 safety.
The company has previously updated its Model Spec, the document that governs how its models behave, to add specific principles around users under 18, and has introduced product-level measures, including parental controls and age prediction tools, alongside a published Teen Safety Blueprint.
OpenAI developed the new policies with input from external organisations, including Common Sense Media, a non-profit focused on technology's impact on children, and everyone.ai.
Related reading
- OpenAI strengthens Sora video safety and consent controls
- OpenAI builds real-time monitor to catch misaligned behaviour in its own AI coding agents
- Mistral AI launches Forge to let companies train their own frontier AI models on proprietary data
Robbie Torney, head of AI and digital assessments at Common Sense Media, said the work was valuable because it helped translate expert knowledge "into guidance that can be used in real systems," adding that content policies of this kind "open the door to broader work on how model behaviour can shape youth-relevant risks over time."
The policies have been published through the ROOST Model Community on GitHub, and the gpt-oss-safeguard model is available via Hugging Face.
The recap
- OpenAI releases prompt-based teen safety policies for developers.
- Policies cover graphic sexual content and dangerous activities.
- Policies are open-sourced via ROOST Model Community GitHub.
