On Tuesday, OpenAI announced the launch of a comprehensive set of prompts aimed at assisting developers in creating safer applications for teenagers. These teen safety policies are designed to integrate seamlessly with OpenAI's open-weight safety model, known as gpt-oss-safeguard.
Instead of starting from scratch, developers can leverage these prompts to strengthen their applications against various issues, including graphic violence, sexual content, and harmful body ideals. The guidelines also address dangerous activities, age-restricted goods, and the portrayal of romantic or violent role play.
These safety policies are structured as adaptable prompts, allowing compatibility with models beyond just gpt-oss-safeguard, though their effectiveness is likely maximized within OpenAI's ecosystem.
In crafting these prompts, OpenAI collaborated with notable organizations focused on AI safety, such as Common Sense Media and everyone.ai. Robbie Torney, Head of AI & Digital Assessments at Common Sense Media, emphasized the importance of these policies, stating, "These prompt-based policies help set a meaningful safety floor across the ecosystem, and because they're released as open source, they can be adapted and improved over time."
OpenAI highlighted that many developers, even those with significant experience, often struggle to translate safety objectives into actionable rules. This gap can lead to inconsistent enforcement and inadequate protection measures. The company noted, "Clear, well-scoped policies are a critical foundation for effective safety systems."
While OpenAI acknowledges that these policies do not resolve all the complexities surrounding AI safety, they build upon previous initiatives, including parental controls and age prediction features. Last year, OpenAI updated its guidelines for large language models to better address interactions with users under 18.
The introduction of these policies marks a significant step forward in fostering a safer digital environment for teens, particularly benefiting independent developers who may lack the resources to establish comprehensive safety measures on their own. As the landscape of AI continues to evolve, such initiatives could pave the way for more robust safety standards and practices in the tech industry.