The collaboration targets the three fundamental components of agentic AI: the reasoning model, the orchestration harness, and the external tools or data sources an agent accesses. By integrating hardware-level detection through NVIDIA BlueField DPUs with TrendAI’s threat intelligence, the partnership seeks to establish a security boundary that exists outside the agent's immediate execution environment, preventing potential bypasses or prompt-based exploits.

Rachel Jin, Chief Platform and Business Officer at TrendAI, emphasized that security must extend beyond simple model alignment to cover every touchpoint within the AI factory. The integrated solution provides visibility into runtime tool calls, enforces strict identity controls, and uses hardware-rooted telemetry to detect lateral movement, ensuring that security remains intact even if a host operating system is compromised. This layered approach is designed to provide organizations with the confidence to deploy autonomous agents without sacrificing data integrity or system safety.