How Falcon ASPM Secures GenAI Applications and Lessons from Dogfooding

This document outlines how CrowdStrike Falcon ASPM secures generative and agentic AI applications—such as Charlotte AI—by continuously mapping complex microservice architectures, monitoring runtime behaviors and dependencies, and assessing risks including vulnerable packages, API access weaknesses, data access segmentation issues, and drift. It highlights a multi-agent architecture with code-level visibility and context-aware risk guidance to detect and mitigate issues like memory poisoning, hallucination propagation, and privilege misuse, emphasizing continuous testing, monitoring, and privacy safeguards within CrowdStrike’s platform.