The Pulse
CrowdStrike to Buy Dreadnode for AI Security Testing
CrowdStrike plans to acquire Dreadnode and add automated AI security testing to its Falcon platform. The tool will probe models, applications and agents before deployment, then send findings to Falcon Guardian for runtime protection.
AI.info Team · Written with AI · fact-checked ·
“Today, at Fal.Con EU 2026, I’m excited to share CrowdStrike’s intent to acquire Dreadnode.”
Michael Sentonas
CrowdStrike announced the proposed acquisition on Nov. 3, 2026, at Fal.Con EU 2026. The company plans to add Dreadnode’s automated AI security testing to its Falcon platform, allowing customers to assess AI systems before putting them into production. CrowdStrike says the tool tests models, applications and agents with known attack techniques to expose weaknesses that teams can fix before deployment.
Testing AI Before Falcon Customers Deploy It
CrowdStrike argues that conventional software security testing misses weaknesses in AI systems because models and agents behave probabilistically and dynamically. The company says traditional testing cannot identify flaws inside the reasoning loops of models and agents. It identifies prompt injection, jailbreaking and agent tool manipulation as risks that can lead to customer data loss, financial loss and reputational damage.
Dreadnode is designed to probe for vulnerabilities before those systems face adversaries. The acquisition would extend the Falcon platform into the layer where businesses build AI, according to CrowdStrike. The company says the goal is to give customers a repeatable way to assess AI systems before deployment.
More Than 50 Ways to Probe an AI System
Users begin by describing what they want to test in plain language. Dreadnode then runs more than 50 attack techniques, including jailbreaks, prompt injection, multi-turn attacks and agent-specific attacks. The tool can test AI systems, models, model repositories, applications, agents and MCP servers before they reach production.
After testing, Dreadnode ranks findings by severity and maps them to the OWASP Top 10 for LLMs, OWASP Agentic Security, MITRE ATLAS and the NIST AI Risk Management Framework. Its assessments combine total findings, attack success rates, severity breakdowns, finding outcomes and other risk metrics into one risk picture. The platform also produces an exportable report for go-or-no-go decisions.
Human Review and Runtime Protection
Human review remains part of the process. Model builders can inspect the attacker prompt and target response, and follow an attack step by step. Operators can reclassify findings, change severity levels and record their reasoning; dashboard metrics update when findings are reclassified.
CrowdStrike says Dreadnode’s findings will feed into Falcon Guardian as policy recommendations, linking weaknesses found during testing to runtime protection. Guardian protects enterprise-developed AI at runtime, while Dreadnode testing data will also support CrowdStrike’s SafeMind models and its research into securing AI systems. The company says Dreadnode will be generally available in Q4 FY27.
The planned addition gives Falcon a predeployment testing function that can inform protection once an AI system is running. CrowdStrike describes its broader aim as securing AI from development through deployment and runtime; Dreadnode’s test results are intended to connect the first stage to Falcon Guardian’s runtime defenses.