CrowdStrike Snaps Up Dreadnode to Stress-Test AI Before It Ships

CrowdStrike is buying Dreadnode. The cybersecurity giant announced a definitive agreement Monday to acquire the automated AI security testing company and bolt its tech onto the Falcon platform, where it will continuously probe every AI model, application and agent for weaknesses before they ever hit production.

The pitch is blunt: AI broke the old testing playbook. Traditional security testing was built for code that behaves the same way every time, but AI models and agents are probabilistic — they can be jailbroken, manipulated through prompt injection, and fail in ways nobody designed a test for. CrowdStrike says those flaws hide inside the model and the agent's reasoning loop, a layer conventional tools never reach.

CEO George Kurtz framed the deal as a trust problem: enterprises are shipping AI faster than they can prove it's safe. '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,' he said, adding the acquisition is meant to let companies 'trust AI and deploy faster, with confidence.'

Dreadnode CEO Brad Palm, whose team will join CrowdStrike's Cyber Superintelligence Lab, said his company was founded to hunt those failures with an attacker's mindset. Dreadnode's technology will also sharpen how CrowdStrike's SafeMind measures the security and cyber capabilities of frontier models.

The automated AI security testing capability is expected to be generally available in the fourth quarter of CrowdStrike's fiscal 2027. The acquisition itself is expected to close during that same quarter, subject to customary closing conditions, with the purchase price paid predominantly in cash plus a stock portion tied to vesting conditions. CrowdStrike made the announcement at Fal.Con Europe 2026 in Barcelona. Financial terms weren't disclosed.