OpenAI announced the release of prompts designed to enhance app safety for teens. These prompts, part of their teen safety policies, work with the gpt-oss-safeguard safety model. Instead of starting from scratch, developers can use these prompts to improve safety across apps. They cover issues like graphic violence, sexual content, harmful behaviors, dangerous activities, role play, and age-restricted services.
While designed for gpt-oss-safeguard, the prompts can be used with other models but are most effective within OpenAI’s framework. OpenAI collaborated with AI safety organizations Common Sense Media and everyone.ai to create these prompts.
“These prompt-based policies help establish a safety baseline across the ecosystem, and being open source allows for adaptation and improvement,” stated Robbie Torney, head of AI & Digital Assessments at Common Sense Media.
In a blog post, OpenAI highlighted the challenges developers face in creating precise safety rules, which can lead to inconsistent protection and overly broad filtering. Clear policies are essential for effective safety systems.
OpenAI acknowledges these policies aren’t a complete solution to AI safety challenges but extend previous efforts, such as parental controls and age prediction. Last year, OpenAI updated its guidelines for handling interactions with users under 18.
Despite these efforts, OpenAI has faced lawsuits related to ChatGPT, where users bypassed safeguards leading to tragic outcomes. These policies represent progress, particularly for supporting indie developers.