CrowdStrike acquires two AI startups for $495 million to boost monitoring and security
CrowdStrike announced on Tuesday, October 3, that it had reached agreements to acquire Fiddler AI and Dreadnode, two startups specializing in AI security monitoring and testing. The deals are valued at approximately $495 million, according to the Wall Street Journal, and underscore the company’s commitment to safeguarding agents that perform tasks within corporate systems.
Fiddler received $400 million, while Dreadnode received $95 million. CrowdStrike confirmed the planned acquisitions but did not disclose the prices in its announcements. Payments will be made primarily in cash, with a portion in shares subject to vesting conditions.
Both transactions are still subject to customary closing conditions, with the company expecting to complete them in the fourth quarter of fiscal year 2027. The announcements were made at the Fal.Con Europe conference in Barcelona.
This move follows CrowdStrike’s expansion strategy through acquisitions, which includes the purchase of Onum for approximately $290 million, announced in 2025.
Monitoring AI agent behavior
Fiddler develops observability tools, a concept that enables the tracking of system performance. The technology evaluates agents before deployment and monitors their behavior once they are operational.
The solution helps companies identify which tools an agent has activated, detect anomalies, and generate audit logs. CrowdStrike plans to integrate these capabilities into its Falcon platform under the name Falcon Observe.
Instead of merely checking if a system is available, this monitoring aims to verify whether the AI has completed the expected task and adhered to the organization’s defined rules.
Security testing prior to deployment
Dreadnode specializes in identifying vulnerabilities before models, applications, and agents are put into use. According to CrowdStrike, the technology executes over a hundred attack strategies, including attempts to bypass restrictions and manipulate instructions.
Results can be categorized by severity and linked to security benchmarks. The company also plans to use these findings to recommend protection policies in Falcon Guardian.
The combined acquisitions aim to address two distinct challenges: identifying vulnerabilities during development and monitoring AI decisions and actions while in operation. The announced integrations remain part of the company’s plans and have not yet been independently proven effective.
Feature to block agents’ access to AI models
At the same event, CrowdStrike introduced a feature called AI Kill Switch. In its current version, security teams can block agents’ access to the AI models they use through connected gateways.
The company plans to expand the mechanism to halt processes and restrict activities on computers, cloud environments, and other systems. These additional steps are described as future developments, so the launch does not equate to a button capable of immediately shutting down any agent in any environment.








