Agent Defense: CrowdStrike Drops $500M to Lock Down Autonomous AI Workflows

Cybersecurity firm CrowdStrike is spending nearly $500 million to acquire two artificial intelligence security startups, expanding its defensive product suite to monitor, test, and protect automated models and software agents. The acquisitions target growing security vulnerabilities as businesses adopt autonomous software agents, open-source models, and vibe-coded applications across their daily operations.

CrowdStrike agreed to acquire Fiddler AI, a startup that builds observability tools for tracking agent behavior inside enterprise networks, along with Dreadnode, a company specializing in automated security testing to detect flaws in models, applications, and agents before deployment. According to reporting from The Wall Street Journal, CrowdStrike is paying roughly $400 million for Fiddler AI and around $95 million for Dreadnode.

The dual acquisitions come as corporate security teams struggle to keep pace with rapid employee adoption of artificial intelligence tools. Workers across every department are testing third-party models, automating complex internal workflows, vibe-coding custom software apps, and deploying armies of autonomous agents to complete routine tasks. This shadow IT activity creates severe security headaches for IT managers, as traditional security software cannot monitor autonomous agent actions, track data connectors, or manage API permission usage effectively.

CrowdStrike already supplies endpoint defense software to thousands of global enterprises and government agencies. The company is moving fast to capture market share in the rapidly growing artificial intelligence defense space. Existing monitoring tools were built to defend static networks, laptops, and cloud servers, but autonomous agents make active choices, execute code, and transfer corporate data independently. Defending modern enterprises requires specialized security tools built to monitor agent behavior, evaluate prompt inputs, and restrict unauthorized background actions in real time.

CrowdStrike had already started building internal detection controls under a project called Falcon Guardian, designed to help IT teams track what software agents do on company laptops and servers. Buying Fiddler AI and Dreadnode speeds up that development roadmap significantly. CrowdStrike plans to integrate Fiddler AI into its core Falcon platform under the name Falcon Observe. The updated tool allows enterprise customers to evaluate agent behavior before deployment, trace real-time execution steps, and feed behavioral telemetry directly into central security dashboards.

Meanwhile, CrowdStrike will integrate Dreadnode directly into the Falcon Guardian product. Dreadnode’s automated red-teaming software tests models and agents for vulnerabilities, feeding test results into Falcon Guardian to generate dynamic security policy recommendations. Combining automated testing with real-time observability gives security administrators clear visibility into agent permissions, preventing autonomous software from leaking sensitive data or opening backdoors into corporate networks.

Founded in 2018, Fiddler AI raised approximately $99 million in venture capital prior to the acquisition, according to PitchBook data. Dreadnode, launched more recently in 2023, raised $14 million from early-stage investors.

As businesses delegate complex tasks to synthetic workers and autonomous software agents, securing model inputs, API connections, and automated outputs becomes a top priority for corporate IT teams. legacy antivirus software and traditional firewalls cannot track decision-making loops inside machine learning software. CrowdStrike’s $500 million investment signals that protecting autonomous software agents will form the next major battleground in enterprise cybersecurity.