CrowdStrike to acquire Fiddler and Dreadnode to bolster AI security

Expanding the Falcon Platform

CrowdStrike has announced plans to acquire two separate companies, Fiddler AI and Dreadnode, to enhance its Falcon security platform. These moves are designed to help enterprises manage the growing risks associated with autonomous AI applications. By bringing these technologies in-house, CrowdStrike aims to provide a more complete lifecycle for AI security.

The company intends to use Dreadnode to provide automated security testing for AI models and applications. Meanwhile, Fiddler will contribute agent observability and explainability features. These tools will help organizations understand how their AI systems behave and why they make specific decisions in production environments.

CrowdStrike stated that the Dreadnode team will join its Cyber Superintelligence Lab. However, the company did not provide specific details on how the Fiddler team will be structured within the organization. Both products are expected to be available to the general public in the fourth quarter of fiscal 2027.

Testing for AI Vulnerabilities

Dreadnode's technology focuses on identifying security weaknesses before an AI system is actually deployed. It allows security teams to test models, applications, and agent reasoning loops against more than 100 different attack strategies. These strategies include common threats like jailbreaks, prompt injections, and multi-turn attacks.

The platform also supports testing Model Context Protocol (MCP) servers. Security teams can use natural language to describe the tests they want to run and then rank the findings based on their severity. This helps teams prioritize which flaws need to be fixed first to ensure the system is safe for use.

CrowdStrike plans to map these vulnerabilities to several established security frameworks. These include the OWASP Top 10 for LLMs, OWASP Agentic Security, MITRE ATLAS, and the NIST AI Risk Management Framework. The results from these tests will eventually feed into Falcon Guardian as specific policy recommendations for users.

Monitoring Autonomous Agents

The acquisition of Fiddler AI addresses a different type of risk: the behavior of AI agents that have not been compromised. CrowdStrike noted that an agent can have valid credentials and approved tools but still make unsupported decisions or ignore instructions. These agents might even claim to have finished work that they never actually completed.

Fiddler’s technology will be offered through Falcon Observe to help companies evaluate agent accuracy and policy compliance. It will trace the reasoning paths and tool calls that an agent takes once it is live. This provides a layer of transparency for how the AI interacts with other software and data.

The system is also designed to detect behavioral drift, which is when an AI starts acting differently than intended over time. It can also identify hallucinations, where the AI generates false information. Additionally, it will maintain records of all agent activity to help companies meet their audit and compliance requirements.

Shared Responsibility and Governance

While these new tools provide better visibility, they do not mean that security teams will be the only ones responsible for AI. Apeksha Kaushik, a senior principal analyst at Gartner, noted that platform and observability teams will still be responsible for the performance and reliability of the systems. Application security and AI teams will still own the secure development and testing phases.

Data and compliance leaders will also remain accountable for the specific use cases and regulatory obligations of the AI. Kaushik suggested that Chief Information Officers (CIOs) should create a clear governance model that defines roles across engineering, risk, and business teams. This ensures that no single department is overwhelmed by the complexities of AI management.

For Chief Information Security Officers (CISOs), the goal is to ensure their teams have the authority to set minimum controls. They need to be able to monitor agent behavior and intervene if the AI creates an unacceptable risk. These acquisitions provide the technical tools to do that, but the organizational structure remains a human responsibility.

Looking Toward Integration

CrowdStrike expects both acquisition deals to close during the fourth quarter of fiscal 2027. The company has aligned the release of the new products with this same timeframe. This timeline is subject to the usual conditions required to finalize a corporate acquisition.

Gartner analyst Apeksha Kaushik cautioned that companies should not view these announcements as final proof of product capabilities. Because the AI threat landscape changes so quickly, the actual integration and features might evolve. She advised customers to wait for more specific details before making long-term plans.

Enterprises are encouraged to ask for clarity on several factors as the deals progress. These include pricing, migration terms, service continuity, and how data will be handled. Customers should also look for information on API access and how they can access historical data changes as the new tools are rolled out.