Qwen-3 Guard Lab: Exploring Automated Prompt Safety for Duke's IT Security

Description

When thousands of students and faculty send prompts to AI systems daily, some inevitably contain sensitive research data, personal information, or content that violates university policies—often without the user realizing it. Duke's IT Security Office (ITSO) wants to explore whether AI-powered safety tools like Qwen-3 Guard could help catch these risks in real-time, but they need to understand what these systems can and can't do before making any decisions. 

That's where you come in: you'll conduct hands-on exploratory research to evaluate Qwen-3 Guard's capabilities and limitations through practical prototyping. You'll build proof-of-concept middleware that intercepts prompts and visualizes safety outputs, design classification schemas to test how well the model aligns with Duke's policy needs, experiment with vector databases for pattern detection, and create dashboard mockups showing how administrators might use such a system. This is deliberately exploratory work—your prototypes are meant to investigate possibilities and surface insights. You'll gain experience with cutting-edge AI safety concepts, learn to evaluate emerging technologies critically, and deliver findings that directly inform ITSO's strategy for responsible AI adoption across campus.

Team

Leaders

Alex Merck

Related Content

Related Links