We don’t know how AI cybersecurity applications will develop over the next five years, nor whether they will privilege attackers or defenders, and yet we must act in favor of positive outcomes. How do we construct an effective observe, orient, decide, act loop under high stakes and significant uncertainty? This talk will explore these themes by extrapolating from the speaker’s experience leading cybersecurity safety assessments and risk mitigations of Meta’s LLM models, by extrapolating from current trends in AI, and while considering the possibility of technological surprise via algorithmic breakthroughs.

Joshua Saxe leads Meta's efforts to integrate security into its large language models (LLMs) and build technologies to protect them from cyberattacks. Before joining Meta, he served as Chief Scientist at Sophos, Principal Investigator on multiple DARPA and NSA-funded cybersecurity research programs at Invincea Labs, and led machine learning security research at Applied Minds. Joshua co-authored the book "Malware Data Science" with Hillary Sanders, published by No Starch Press. He has authored dozens of scientific papers and patents on security AI.