OpenAI adds open source tools to help developers build for teen safety

OpenAI announced the release of a new set of open-source safety prompts aimed at helping developers create AI applications that are safer for teenage users. Developed in collaboration with AI safety organizations Common Sense Media and everyone.ai, these prompt-based policies are designed to address specific high-risk areas for teens, including graphic violence, sexual content, harmful body ideals and behaviors, dangerous activities and challenges, romantic or violent role-play, and age-restricted goods and services. The policies are formatted as prompts, making them easily compatible with various AI models, though they are optimized for use with OpenAI's own open-weight safety model, gpt-oss-safeguard.

OpenAI stated that the initiative addresses a common challenge where developers, even experienced teams, struggle to translate broad safety goals into precise, operational rules, which can lead to protection gaps or inconsistent enforcement. The company emphasizes that these prompts are not a complete solution to AI safety but build upon previous efforts like parental controls and its Model Spec guidelines for users under 18. The release comes as OpenAI faces scrutiny over teen safety, including lawsuits related to extreme ChatGPT use. By open-sourcing these policies, OpenAI aims to set a "meaningful safety floor" across the ecosystem that indie developers and larger teams can implement and adapt over time.