Safety of GenAI Through the Lens of Security and Cryptography

Explore a comprehensive lecture by Somesh Jha from the University of Wisconsin-Madison examining the safety challenges of Generative AI through security and cryptographic perspectives. This 1-hour 6-minute talk, presented at the Simons Institute as part of the "Theoretical Aspects of Trustworthy AI" series, delves into the intersection of traditional security frameworks and modern AI systems. Learn how cryptographic principles can be applied to address emerging safety concerns in generative models and gain insights into theoretical approaches for building more trustworthy AI systems.