At Fal.Con Europe 2026 in Barcelona, cybersecurity firm CrowdStrike announced two linked initiatives designed to curb the risks posed by autonomous AI agents in corporate environments. The company debuted an AI “kill switch” that can block an agent’s access to its underlying models through a gateway, and disclosed a pending acquisition of AI‑observability startup Fiddler AI, valued at roughly $400 million according to The Wall Street Journal. Together, the tools are intended to give security teams the ability to stop rogue behavior and to understand why an agent acted the way it did.

How the AI kill switch works

CrowdStrike’s kill switch operates at the AI gateway level, intercepting requests from an agent to the model that powers its decisions. When an administrator triggers the switch, the gateway denies the agent’s access, effectively pausing its activity while an audit record is created. The initial release does not terminate processes on endpoints, but the company said future updates will allow direct endpoint termination and coordinated restrictions across cloud and on‑premise systems. The feature is integrated into the Falcon Guardian platform and can be applied to agents built on tools such as Claude Code, OpenAI Codex, GitHub Copilot, Microsoft 365 Copilot, Cursor and OpenCode.

Fiddler AI acquisition adds observability

In parallel with the kill switch launch, CrowdStrike signed an agreement to acquire Fiddler AI, a startup that specializes in monitoring AI models and explaining their behavior. The deal, expected to close in the fourth quarter of fiscal 2027, is being paid primarily in cash with a portion of stock subject to vesting. Fiddler’s software evaluates agents before deployment, records prompt‑to‑action traces, and tracks changes in behavior once the agents are live. This observability layer is meant to complement the kill switch by providing context on what an agent was doing before it was halted, helping security teams investigate incidents and auditors verify compliance.

Implications for enterprise security

The combined offering targets a growing concern among enterprises: autonomous agents can execute code, access sensitive data and interact with business applications without continuous human oversight, raising the potential for costly mistakes or malicious exploitation. Prior incidents, such as a vulnerability in Anthropic’s Claude Code GitHub Action that could expose credentials, highlight the need for both immediate containment and post‑event forensics. By integrating a kill switch with detailed decision‑trace capabilities, CrowdStrike aims to give organizations a two‑pronged defense—rapid interruption of harmful actions and a clear audit trail for root‑cause analysis.