CrowdStrike Advances AI Security and Safety from Endpoint to Cloud

AI agents are transforming enterprise computing. What began with employees using AI assistants is evolving into autonomous agents that access enterprise resources, invoke tools, execute code, and take action at machine speed.

This introduces a new security and safety challenge. Organizations must protect agents from attacks while ensuring those agents operate safely and as intended. Security teams need to discover where agents operate, understand what they were instructed to do, trace their actions, and intervene when their behavior becomes dangerous.

Traditional security approaches provide only part of the picture. Endpoint runtime security monitors execution but lacks the AI-specific context needed to explain what triggered it. Network inspection can monitor AI communications but not the actions that agents take on systems. Agent-level integrations can inspect AI interactions but cannot reveal downstream execution.

At Fal.Con Europe 2026, CrowdStrike unveiled innovations that advance AI detection and response (AIDR) across three critical dimensions: architecture, coverage, and controls. We also introduced a new AI kill switch for CrowdStrike Falcon® Guardian that empowers security teams to rapidly cut off AI model access at AI gateways when agents become dangerous. And we shared the general availability of CrowdStrike Falcon® Complete for Guardian, which extends 24/7 managed detection and response to AI security.

The Right Architecture for Securing AI Agents

Securing AI agents requires visibility and control across three critical layers: OS-level runtime execution, network communications, and agent interactions. Falcon Guardian brings these layers together to connect what agents are instructed to do with what they execute.

OS-level runtime security is the foundation (Available now): The Falcon sensor provides deep visibility into agents’ activity, including the processes they launch, commands they run, and resources they access. This visibility is not achievable directly via network inspection or per-agent hook integration approaches.

Endpoint network inspection extends visibility into agent communications (Pre-beta): CrowdStrike announced plans to expand this capability to macOS, where it will inspect traffic entering and leaving the endpoint to help security teams identify risky interactions.

Agent-level hooks provide direct inspection and policy enforcement within supported AI agents (Available now): Hooks for Claude Code, OpenAI Codex, GitHub Copilot, OpenCode, and Cursor enable organizations to inspect prompts and block malicious activity at the source, before harmful actions are executed.

Together, these three layers give Falcon Guardian visibility into agent instructions, communications, and runtime execution. CrowdStrike® Agent Graph connects this activity, linking what agents were instructed to do with what they actually executed and the resources they affected. This gives security teams the context to understand agent behavior, investigate threats, and trace their impact from prompt to process.

See it in action:

The Right Coverage: Secured AI Wherever It Operates

Falcon Guardian initially focused on securing workforce AI adoption and autonomous agents at the endpoint, where employees interact with AI and agents execute with user permissions.

Now, as enterprises build and deploy their own agents in production environments, CrowdStrike is expanding Falcon Guardian's AI-specific protections across more of the cloud infrastructure the CrowdStrike Falcon® platform already secures. Newly announced capabilities will extend Falcon Guardian coverage across Linux, cloud workloads, Kubernetes, and managed AI platforms.

Expanded cloud coverage (Pre-beta): On Linux, Falcon Guardian will bring agent discovery, visibility, and AI-specific detections to Claude Code and OpenAI Codex. In cloud workloads and Kubernetes, Falcon Guardian will extend these capabilities to enterprise-developed AI agents and applications, including those built with LangChain and LangGraph. Using the Falcon Linux sensor, deployed directly or as a Kubernetes DaemonSet, Falcon Guardian will connect AI activity with surrounding runtime, workload, and identity context.

For provider-managed AI environments, Falcon Guardian will expand agent discovery and AI-layer visibility, beginning with Amazon Bedrock AgentCore. This will help security teams identify managed agents, understand their components and interactions, and assess AI-specific risks without requiring access to provider-controlled infrastructure.

Claude Cowork integration (Pre-beta): CrowdStrike also announced a new Anthropic integration that extends Falcon sensor protection inside the virtual machine where Claude Cowork executes, bringing runtime security directly into the agent's execution environment.

Falcon AI Gateway (Available now): CrowdStrike announced the general availability of Falcon AI Gateway, which provides centralized policy and enforcement for supported AI traffic to extend protection beyond environments with direct sensor coverage. Falcon AI Gateway is included with Falcon Guardian. 

The Right Controls: Detect Threats and Govern AI

Comprehensive visibility must be paired with the ability to detect threats, enforce policy, and stop agents that become compromised or behave dangerously.

Strengthen AI threat detection with SafeMind (Available now): Falcon Guardian's AI-specific threat detection is now powered by CrowdStrike SafeMind, a family of purpose-built AI security models developed by the CrowdStrike Cyber Superintelligence Lab in partnership with NVIDIA. These models help identify malicious intent in AI interactions, including prompt injection and other attacks targeting AI systems.

Falcon Guardian expands AI supply chain protection by discovering malicious AI skills and MCP servers to help security teams identify threats before they enter the AI environment. Upcoming enhancements will also introduce detection of agentic intent misalignment, helping identify when agents act outside the intent of their instructions.

Identify agent insider threats through behavioral baselining (Pre-beta): CrowdStrike Falcon® Next-Gen SIEM is gaining agent insider threat detection. By establishing behavioral baselines for agents and their owners, it can identify meaningful deviations even when an agent operates within its granted permissions. Its correlation with Falcon Guardian detections and broader Falcon telemetry helps security teams recognize potential compromise or misuse.

Stop rogue agents with an AI kill switch (Available now): Security teams need the ability to intervene quickly when AI agents become compromised or behave dangerously. CrowdStrike introduced a new AI kill switch, now available to Falcon Guardian customers, that enables organizations to rapidly cut off AI model access through centralized gateway enforcement, helping stop rogue agent activity before it causes further harm.

Extend 24/7 managed protection with Falcon Complete for Guardian (Available now): CrowdStrike also announced the general availability of Falcon Complete for Guardian, extending managed detection and response (MDR) services to AI security. With 24/7 monitoring, investigation, and response from CrowdStrike's expert security team, customers can extend managed protection to AI-specific threats without adding operational complexity.

Additional enhancements to manage AI security at scale (Available now): Falcon Guardian also introduces enhancements to help organizations manage AI security at scale, including Falcon Data Replicator integration for prompt activity, Falcon Flight Control support, new executive summary and token consumption dashboards, and expanded Windows AI application coverage.

One Platform for AI Security and Safety

Organizations need AI-specific protection connected to the runtime, identity, cloud, and security operations capabilities they already rely on to secure autonomous agents and defend the enterprise.

CrowdStrike delivers that unified approach through the Falcon platform. From employee AI adoption to autonomous agents operating across the enterprise, Falcon Guardian helps organizations discover and govern AI, understand its actions, detect threats, and stop dangerous behavior. By extending AI security across the Falcon platform, CrowdStrike helps organizations secure AI from adoption to action and innovate with confidence.

Learn more about CrowdStrike Falcon Guardian.

Disclaimer

This blog includes discussion of unreleased services or features. Any unreleased services or features referenced here are still in development and subject to change. Customers should make their purchase decisions based upon features that are currently available.