SI Agent News
DevelopingClaudeAgentsSecurityBusiness

CrowdStrike to acquire Fiddler AI and Dreadnode to bolster AI security

The cybersecurity giant is expanding its toolkit to help companies monitor, test, and shut down autonomous AI agents.

98%Reduction in AI evaluation costs using Fiddler

The short version

Why it matters

As companies deploy autonomous AI agents to perform tasks, they need a way to audit those actions for safety and compliance. These acquisitions provide the 'flight recorder' and 'stress test' necessary to manage the risks of unpredictable AI behavior.

Expanding the AI Security Suite

CrowdStrike has reached an agreement to purchase two specialized AI security startups, Fiddler AI and Dreadnode. The company shared the news during its Fal.Con Europe conference in Barcelona. These moves signal a major push into the infrastructure required to manage autonomous AI agents safely.

The deal structure involves a combination of cash and stock with specific vesting conditions. While CrowdStrike did not reveal the exact purchase price, both acquisitions are scheduled to close in the fourth quarter of fiscal 2027. This timeline allows the company to integrate the technologies into its existing platform over the next few years.

By bringing these companies into its fold, CrowdStrike aims to address a growing corporate concern regarding AI transparency. As the company noted, simply blaming a model for an action is no longer a sufficient answer for regulators or corporate boards. These tools are designed to provide the specific evidence needed to explain AI behavior.

The Flight Recorder for AI

Fiddler AI will be rebranded as Falcon Observe and integrated into the Falcon platform. Its primary role is to act as a flight recorder for AI agents, documenting every action the agent takes. This level of observability is critical for companies that need to prove their AI systems stayed within their assigned tasks.

Fiddler's technology uses purpose-built models to analyze these actions. According to the company, these models can reduce the costs associated with evaluating AI performance by up to 98%. This efficiency helps large enterprises scale their AI usage without incurring massive overhead for monitoring and auditing.

Krishna Gade, the co-founder and CEO of Fiddler, stated that the goal is to help enterprises answer if they can trust what their agents are doing. By providing a clear trail of logic, the tool helps move AI from a 'black box' to a transparent system that can be audited for compliance and safety.

Stress-Testing Against Attacks

The second acquisition, Dreadnode, focuses on proactive defense by finding weaknesses before they are exploited by malicious actors. The Dreadnode team will join CrowdStrike’s Cyber Superintelligence Lab to help build tools that simulate various attack strategies. These include common threats like jailbreaks and prompt injections.

Dreadnode's software can run more than 100 known AI attack strategies against models, applications, and agent reasoning loops. It maps these findings to established security frameworks, such as the OWASP Top 10 for large language models and the NIST AI Risk Management Framework. This helps organizations standardize how they measure and mitigate AI risks.

Brad Palm, the CEO of Dreadnode, noted that AI fails in ways that traditional testing was not designed to catch. Because AI vulnerabilities are unique, the company uses a team that thinks like an attacker to find flaws. This proactive approach ensures that models are hardened against manipulation before they go live in production environments.

Immediate Tools and Integrations

While the acquisitions will take time to fully integrate, CrowdStrike has already released a new tool called the AI Kill Switch. This feature is available now within the Falcon Guardian product. It allows a security administrator to immediately stop an agent if it begins to behave in an unauthorized or 'rogue' manner.

The Kill Switch works by stopping the agent at the gateway and blocking it from launching again. It also cuts off the agent's access to the company's internal network. Every action taken by the administrator when triggering the switch is logged to ensure a clear record of the intervention.

Additionally, CrowdStrike is expanding its reach by bringing Falcon Guardian into Anthropic’s Claude Cowork. The sensor runs inside the virtual machine where agents execute code. This integration is currently in beta on macOS, with Windows support expected to arrive in late November.

What to watch

Try it

Check your company's AI policy to see if there is a plan for 'human-in-the-loop' oversight or a protocol for shutting down autonomous agents.

Sources and links

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

Listen to the AI Wire headlines

Listen to the AI Wire

Listen to our daily AI shows every morning: on your commute, in the shower, wherever you are

Listen to the daily shows

First reported by The Next Web · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

More AI news