David Hollingworth 
CrowdStrike has unveiled a new Secure-by-Design AI Blueprint developed with NVIDIA, aimed at embedding cyber security protections directly into the runtime layer used to build and operate autonomous AI agents.
The blueprint integrates the CrowdStrike Falcon platform into NVIDIA OpenShell, an open-source runtime environment designed to apply policy guardrails to AI agents as they execute tasks.
The architecture is intended to give organisations governance, monitoring, and control across the entire AI life cycle, from development through to production deployment.
According to CrowdStrike, traditional security controls were designed for static systems and human users, making them poorly suited to govern autonomous software operating at machine speed. The Secure-by-Design blueprint aims to address that gap by embedding continuous security enforcement directly into the AI agent stack.
Under the architecture, the Falcon platform integrates with the OpenShell runtime – part of NVIDIA’s Agent Toolkit – to provide isolated execution environments with built-in policy enforcement and private inference capabilities. This allows security controls to be applied wherever agents run, including local deployments on GPU-accelerated systems such as NVIDIA DGX Spark and NVIDIA DGX Station, as well as cloud-based deployments built on NVIDIA’s open-source AI-Q framework.
The platform enables real-time monitoring of agent behaviour, including prompts, responses and actions, with the goal of preventing prompt manipulation, constraining unsafe behaviour and enforcing policy across AI workloads.
Daniel Bernard, chief business officer at CrowdStrike, said the shift to autonomous systems fundamentally changes the security landscape.
“As we enter the agentic era, agents no longer simply assist – they act,” Bernard said in a statement.
“This shift fundamentally changes the security equation, and security must be embedded into the AI stack itself. Together with NVIDIA, we are delivering a secure-by-design architecture that enables organisations to operationalise agents with confidence and control.”
Justin Boitano, vice president of enterprise platforms at NVIDIA, said the partnership is intended to help enterprises safely deploy AI agents at scale.
“Autonomous agents will fundamentally reshape how we work,” Boitano said.
“By integrating CrowdStrike’s security platform with the NVIDIA Agent Toolkit, we’re enabling enterprises to build and scale safer autonomous AI agents to transform operations and securely generate intelligence at the speed of business.”
The blueprint also introduces identity-driven governance capabilities designed to ensure AI agents operate within defined privilege boundaries when accessing data, APIs and enterprise services.