Using LLMs to Find and Prioritize Vulnerabilities Is No Easy Task

Despite the promise of large language models (LLMs) in streamlining vulnerability detection, they're currently riddled with high false-positive rates and struggle to understand the context of security scans. This means AppSec experts still have to spend extra time sifting through incorrect data, highlighting the need for human oversight. The implications are significant as it underscores the ongoing challenge in balancing automation with the nuanced, context-aware expertise that humans provide, especially in the ever-evolving landscape of application security.