CrowdStrike unveils Falcon Guardian to secure AI agents

As reported by Silicon Angle, CrowdStrike has introduced Falcon Guardian, a new software designed to identify and disable unapproved artificial intelligence agents operating within a company's systems. The announcement was made at the company's Fal.Con 2026 conference in Las Vegas.

Falcon Guardian builds upon previous offerings by adding enforcement capabilities. The software aims to provide a comprehensive view of AI agent activity by monitoring the endpoint where agents process information and execute code. The Falcon sensor inventories both known and shadow AI agents across Windows and macOS, logging their deployment and security status. Guardian then links agent behavior to endpoint telemetry, tracing actions from user prompts through the identity, tools, and skills used, to the impact on the system.

Administrators can define approved agents, with any unlisted agents being blocked. The system also detects attacks targeting AI agents and instances of agents behaving improperly, reconstructing the execution chain and assessing the potential damage in real-time. Founder and CEO George Kurtz stated that the endpoint approach, pioneered by CrowdStrike for EDR, is necessary for AI security due to the increased speed of attacks. Future components include AI Gateway for policy enforcement on AI traffic and Falcon Complete for Guardian, a managed detection and response service. Agent telemetry will be integrated into Falcon Next-Gen SIEM for correlation with other security data.

Source: Silicon Angle