TRUSTWORTHY GENERATIVE AI

Catalog description

This course covers the landscape of adversarial attacks on generative AI and their implications for security and privacy. Combining foundational research with cutting-edge developments, students explore key themes in security, robustness, and alignment, tracing their roots from classical machine learning theory to modern large-scale models. Emphasis is placed on understanding not just what methods succeed or fail, but why, fostering the analytical skills needed to evaluate current defenses and craft principled, trustworthy AI systems. The course also covers mitigation techniques and the real-world challenges of securing generative AI deployments. (3 credits) fall