SAN JOSE, California, 16 March 2026 — CrowdStrike unveiled a “Secure-by-Design AI Blueprint” with NVIDIA today. This protects the growing agentic AI in cybersecurity market. Announced at GTC, the architecture integrates CrowdStrike Falcon with NVIDIA OpenShell. This secures autonomous agents across cloud and local sites.
The blueprint provides a standardized framework for building agents. These agents can reason and act independently. Moreover, they also maintain strict security guardrails. By embedding protection in the AI runtime, companies prevent prompt manipulation. This also stops unauthorized data access. These actions occur at machine speed.
This collaboration addresses the unique risks of the Agentic AI era, where autonomous systems operate as privileged identities with access to critical applications. Traditional security measures often fail to govern these dynamic systems, necessitating the continuous, policy-based enforcement provided by this new architecture.
Key features include AI Policy Enforcement for real-time protection. Additionally, Identity-Based Governance limits access to sensitive APIs. Consequently, these tools secure every agent action. Furthermore, the blueprint uses Falcon Cloud Security to monitor deployments. Specifically, this covers NVIDIA’s AI-Q research framework. Moreover, it protects local DGX Spark systems.
Industry leaders emphasized that as AI shifts from assistive “copilots” to autonomous agents, security must be native to the AI stack. CrowdStrike Chief Business Officer Daniel Bernard noted that this shift changes the security equation, requiring agents to be “secure by design” to ensure operational confidence.
The initiative positions CrowdStrike at the forefront of the agentic AI in cybersecurity market by offering unified visibility across the entire AI lifecycle. By partnering with NVIDIA, the company provides enterprises with the infrastructure needed to scale autonomous operations without compromising safety or data integrity.
