FlagThis

Abstract

This talk explores the emerging capability of large language models (LLMs) to autonomously identify and exploit zero-day vulnerabilities in critical software, such as the Linux kernel and popular CMS platforms. The speaker argues that the historical balance between attackers and defenders is collapsing as LLMs lower the barrier for high-end vulnerability research. Key takeaways include the demonstration of autonomous exploit generation and a warning that the exponential growth of AI capabilities necessitates an immediate, industry-wide shift in defensive strategies.