The Pulse
CrowdStrike Adds an AI Agent Kill Switch to Falcon Guardian
CrowdStrike has introduced AI Kill Switch in Falcon Guardian, where security teams can block AI model access through connected gateways. Coordinated containment across endpoints, cloud, identities, SaaS and browsers is planned as a later ex
AI.info Team · Written with AI · fact-checked ·
CrowdStrike’s AI Kill Switch can block an agent’s access to AI models now; coordinated containment across endpoints, cloud, identities, SaaS and browsers is a planned extension. The distinction matters: the control available in Falcon Guardian acts through AI gateways, while the company describes broader enforcement across its platform as coming soon. For security teams, the launch is an immediate way to interrupt model access, not the full cross-system response CrowdStrike says it is building.
The company frames the problem as an agent’s ability to set off actions across credentials, applications, files and enterprise systems. Its proposed response starts with cutting off the models that power an agent’s decisions and actions, then aims to reach the other systems where agents operate.
Falcon Guardian blocks model access through gateways
Security teams activate AI Kill Switch in Falcon Guardian, which directs connected AI gateways to block access to the AI models that power agents’ decisions and actions. Those gateways can be CrowdStrike’s Falcon AI Gateway or a third-party gateway already in use. The mechanism therefore depends on a connected gateway to interrupt an agent’s access to its models.
Falcon AI Gateway is generally available in Falcon Guardian and is included at no additional cost with Falcon Guardian. CrowdStrike says it extends AI security to applications and workloads beyond direct endpoint sensor coverage, including unmanaged environments. The gateway also centralizes model access, security policies, credential management, and usage and cost controls within the Falcon platform.
Broader containment is still in development
CrowdStrike says it will extend the kill switch beyond gateway enforcement to coordinate containment across the Falcon platform. The planned controls cover several points where agents may operate: endpoints, cloud environments, user and agent identities, SaaS applications and browser sessions. On endpoints, the company describes terminating rogue agents and associated processes, then preventing them from restarting.
For cloud agents, the planned response includes containment or termination. For identities, CrowdStrike says Falcon Next-Gen Identity Security would revoke agent permissions at runtime and restrict access to enterprise resources. Falcon Shield would restrict or terminate SaaS activity, while Seraphic Enterprise Browser would terminate sessions and disrupt unauthorized activity through web applications.
Agent Graph links intent to execution
The company says CrowdStrike Agent Graph connects an agent’s intent to its downstream execution. It is designed to show what an agent was asked to do, what it did, and which systems or resources were affected. That context, CrowdStrike says, can help security teams determine where to intervene.
CrowdStrike Charlotte Agentic SOAR is described as the no-code orchestration and response layer for coordinating enforcement across the Falcon platform. The company says it can support automated AI Kill Switch workflows triggered by AI-specific detections or risky agent behavior. Its stated aim is to let security teams stop individual rogue agents or groups of agents while minimizing disruption to business operations.
CrowdStrike flags unreleased features as subject to change
The announcement distinguishes the available gateway control from the broader enforcement features still in development. CrowdStrike’s disclaimer says unreleased services or features are subject to change and advises customers to base purchase decisions on features that are currently available. The current offering described in the post is the AI Kill Switch in Falcon Guardian, using connected gateways to block model access.