CrowdStrike acquires two AI security firms and adds kill switch

Expanding the AI security portfolio

CrowdStrike has announced the acquisition of two specialized AI security startups to strengthen its defenses against modern cyber threats. The company is moving to reinforce the security foundations required for the growing AI era, where hacking exploits targeting artificial intelligence are becoming increasingly common. By bringing these firms into its fold, CrowdStrike aims to offer more comprehensive protection for enterprise AI systems.

The acquisitions include Fiddler AI, a company that creates observability software. This software is designed to inspect and analyze the specific actions taken by AI agents within a corporate environment. By understanding these actions, companies can better monitor how their automated systems are interacting with data and other software. This level of visibility is becoming a priority as businesses move toward more autonomous workflows.

The second acquisition is Dreadnought, which operates a platform focused on proactive security. Dreadnought's technology automatically identifies security vulnerabilities before a company actually deploys its AI models or agents. By finding these flaws early in the development cycle, organizations can prevent many types of attacks before they ever reach a live production environment. CrowdStrike did not disclose the official purchase prices, but reports suggest the deals are worth hundreds of millions of dollars.

Integrating new technologies

CrowdStrike plans to merge the capabilities of these two startups into its existing Falcon Guardian platform. Falcon Guardian is already used by enterprises to detect the activity of AI agents in real time and protect them from external attacks. The integration of Fiddler AI's software will provide a more detailed layer of oversight for these agents. This allows for a more holistic view of how AI is functioning across a network.

The technology from Fiddler AI is described as a way to see inside the 'black box' of AI behavior. According to CrowdStrike CEO George Kurtz, the software provides observability that can explain exactly what an agent did and why. This level of transparency is intended to help administrators understand the reasoning behind an AI's actions. Without this clarity, it is difficult for security teams to ensure that an agent is following company policy or safety guidelines.

Dreadnought's automated vulnerability detection will also be woven into the Falcon Guardian platform. This means that the security checks performed by Dreadnought can be standardized across the CrowdStrike ecosystem. By automating the search for flaws, the platform helps reduce the manual labor required by security engineers. It ensures that the AI models being used by a company are structurally sound and less susceptible to being hijacked by malicious actors.

The AI kill switch feature

In addition to the acquisitions, CrowdStrike is launching a specific 'kill switch' feature for its security products. This feature is designed to act as an emergency brake for AI agents that begin to behave in risky or unpredictable ways. If an agent performing corporate tasks goes out of control, the kill switch provides a method to stop it immediately. This is a critical safety measure for companies that are worried about the autonomy of their AI tools.

The kill switch works by allowing administrators to quickly cut off the network access of a specific AI agent. Once the network connection is severed, the agent can no longer operate or communicate with other systems. This prevents a malfunctioning or compromised agent from causing widespread damage or leaking sensitive information. The feature is built directly into the Falcon Guardian platform to ensure it is accessible to security teams during an incident.

This functionality addresses a major concern for enterprise leaders who are hesitant to give AI agents full autonomy. While AI can increase efficiency, the possibility of an agent being hacked or behaving erratically presents a significant risk. The kill switch provides a layer of human oversight, ensuring that a person can always intervene. It serves as a fail-safe to mitigate the risks associated with the rapid deployment of automated AI systems.

Addressing the rise of AI hacking

The move by CrowdStrike comes at a time when hacking that exploits AI is running rampant across the globe. Cybercriminals are finding new ways to target machine learning models and the agents that interact with them. Because of this trend, major cybersecurity firms are accelerating their merger and acquisition activities. They are racing to build the necessary security infrastructure to protect the next generation of computing.

As more companies adopt AI to handle complex tasks, the attack surface for these organizations expands. An AI agent might have access to internal databases, customer information, or sensitive communications. If that agent is not properly secured, it can become a gateway for attackers to enter a company's network. CrowdStrike's strategy is to create a multi-layered defense that covers the entire lifecycle of an AI model.

By combining observability, vulnerability detection, and emergency controls, CrowdStrike is attempting to create a standard for AI safety. The goal is to move beyond simple firewalls and toward a system that understands the specific behaviors of AI. This approach helps companies adopt AI technology more confidently by reducing the uncertainty surrounding how these systems might fail or be manipulated by outside forces.

What this means for enterprises

For businesses, these developments mean that AI security is becoming a more specialized and sophisticated field. Companies will likely see more integrated tools that offer both proactive scanning and real-time monitoring. Instead of using separate tools for different security needs, they can use a single platform like Falcon Guardian to manage their AI risks. This consolidation can simplify the job for IT departments that are already stretched thin.

The focus on 'observability' is particularly important for regulated industries that must explain how their systems make decisions. If an AI agent makes a mistake or performs an unauthorized action, the company needs to be able to audit that event. The technology from Fiddler AI helps provide that audit trail, making it easier to comply with internal policies and external regulations regarding automated decision-making.

Looking ahead, the integration of these tools suggests that AI security will become a standard requirement for any enterprise AI deployment. As the technology matures, the ability to quickly shut down a rogue agent will become as common as a standard firewall. CrowdStrike's actions signal a shift toward a more controlled and supervised environment for corporate AI, where safety and transparency are prioritized alongside performance.