CrowdStrike to acquire AI security tester Dreadnode as agent risks mount
What's the deal? CrowdStrike has signed a definitive agreement to acquire Dreadnode, a startup focused on automated AI security testing. The deal, announced November 3 at Fal.Con Europe in Barcelona, will be paid predominantly in cash, with a portion in stock subject to vesting conditions. Terms were not disclosed.
What each side brings: Dreadnode runs more than 100 known attack strategies — jailbreaks, prompt injection, and agent-specific attacks — against AI models, applications, and agent reasoning loops. CrowdStrike plans to fold that into its Falcon platform, ranking findings by severity and mapping them to frameworks such as the OWASP Top 10 for LLMs and MITRE ATLAS.
Why now? Enterprises are deploying AI faster than they can secure it, and conventional testing — built for code that behaves the same way every time — struggles with probabilistic models. "A model can be jailbroken, an agent can be manipulated, and most organizations have no way to find those flaws until the AI is already in production," said chief executive officer and founder George Kurtz.
What's the endgame? The acquisition fills a gap between catching weaknesses before deployment and stopping risky behaviour in production. Dreadnode's findings will feed Falcon Guardian, CrowdStrike's runtime agent security tool launched in September, while its testing data will strengthen CrowdStrike's SafeMind models.
The Dreadnode team will join CrowdStrike's Cyber Superintelligence Lab. "We founded Dreadnode to expose those weaknesses before AI reaches production," said Brad Palm, Dreadnode's chief executive officer.
What could go wrong? The product is some way off: automated AI security testing will not be generally available until CrowdStrike's fourth quarter of FY27, when the deal is also expected to close. Integration and customer adoption remain unproven.
The signal: The deal reflects a crowded race to secure the full AI lifecycle. Palo Alto Networks has been expanding its Prisma AIRS product with agent discovery, red teaming, and runtime controls, and CrowdStrike itself announced new controls and an AI "kill switch" in November. As agentic enterprises scale, pre-deployment testing is becoming a battleground for cybersecurity vendors.
Read more: AP News
Image credit: Idaho National Laboratory
