OpenAI has introduced prompt-based safety policies aimed at enhancing AI systems to protect teenagers from inappropriate content. These policies are designed to integrate with the open-weight model gpt-oss-safeguard, offering developers practical classifiers that meet safety requirements for real-world applications. They specifically address risks such as graphic violence, sexual content, harmful body image issues, and age-restricted goods. The framework, developed in collaboration with organizations like Common Sense Media, enhances clarity and consistency in teen safety measures. It provides tools for real-time filtering and offline content analysis while allowing open-source adaptability through the ROOST Model Community for various languages and use cases. This initiative is a foundational step towards layered safeguards, highlighting the importance of ongoing refinement in AI safety policies. Want to explore more about AI, technology, and digital diplomacy? Engage with our Diplo chatbot for insights!
Source link
Share
Read more