CrowdStrike is moving to police the AI agent era from end to end. At its Fal.Con Europe 2026 conference in Barcelona on Tuesday, the cybersecurity company unveiled a new AI Kill Switch for its Falcon Guardian product and announced two acquisitions aimed at watching agents before they launch and while they run.
The Wall Street Journal reported the two deals are worth a combined $495 million: roughly $400 million for Fiddler AI and about $95 million for Dreadnode, paid mostly in cash with a portion in stock subject to vesting. CrowdStrike itself did not disclose purchase prices, and both transactions are expected to close in its fourth quarter of fiscal 2027, subject to customary conditions.
A kill switch for rogue agents
The headline feature is Falcon Guardian's new AI Kill Switch, which lets administrators terminate agents that start acting abnormally. Guardian secures agents where they execute, on the endpoint at runtime, and the update unifies three control points: agent-level hooks that enforce what agents are told to do, endpoint network inspection that sees when they communicate with models and tools, and the endpoint itself, which stops unauthorized activity.
CrowdStrike is also extending Guardian beyond the endpoint to cloud workloads, software-as-a-service apps, virtualized agents including Claude Cowork, and cloud agent frameworks including AWS AgentCore. The pitch is straightforward: agents now hold real tool access and permissions, so a misbehaving agent can cause damage far faster than a misbehaving employee, and security teams want a single button to cut it off.
What Fiddler and Dreadnode add
Fiddler AI, founded in Palo Alto in 2018, builds observability and explainability software for AI systems. Its technology will become a product called Falcon Observe, generally available in the fourth quarter of fiscal 2027. It traces an agent's reasoning paths and tool calls, detects behavioural drift, and tracks token usage, letting enterprises see what an agent did, why it acted, and whether it stayed within its task. CrowdStrike says the product can cut evaluation costs by up to 98 percent by using purpose-built models instead of external large-language-model calls, and it supports frameworks including LangGraph, Amazon Bedrock and Google's Agent Development Kit through OpenTelemetry.
Dreadnode, founded in 2023, takes the other end of the lifecycle: automated security testing of AI models, applications and agents before they reach production, probing for weaknesses that could let users bypass safety restrictions. Its findings will feed Falcon Guardian as policy recommendations, so a flaw found in testing strengthens runtime protection, and its team will join CrowdStrike's Cyber Superintelligence Lab.
Why CrowdStrike is buying the whole lifecycle
Taken together, the pieces sketch a full lifecycle for agent security: Dreadnode tests an agent before launch, Fiddler watches it in production, and the kill switch shuts it down if things go sideways. CrowdStrike is pitching the bundle to companies facing auditors, regulators and boards asking what their agents actually did — a question point tools answer only in fragments.
The strategy also deepens CrowdStrike's push into AI detection and response, a category it has been building as enterprises rush agents into customer service, coding, finance and operations. The open question is adoption: the acquired capabilities do not reach general availability until late in the company's fiscal 2027, and customers will judge whether one bundled platform beats best-of-breed tools for testing, observability and runtime control. For now, though, CrowdStrike has staked its claim — the agent economy needs a flight recorder, a test lab and an emergency brake, and it intends to sell all three.



