TrendAI, the AI security business unit of Trend Micro Incorporated, has announced support for the NVIDIA Agent Safety Platform, the chipmaker's full-stack infrastructure for running autonomous AI agents safely at enterprise scale. The announcement, issued on Monday, September 28, via PR Newswire and republished across Asia on September 29, positions TrendAI's Vision One platform as the threat-intelligence and detection layer on top of NVIDIA's enforcement stack. The goal: help enterprises move their AI agent fleets from pilot projects into full production without losing control.

Intelligence on Top of Enforcement

According to NVIDIA, a component called OpenShell enforces security policy outside the agent's execution environment, where it cannot be prompted away or bypassed. NVIDIA's BlueField-4 data processing units (DPUs) and DOCA software add a host-independent security layer implemented in silicon. TrendAI says it complements that enforcement layer with the pieces around it: the intelligence that informs what the policy should be, the visibility to see what agents are doing, and the detection and response capabilities to act on it.

From Pilot Fleets to Production


The integration centers on TrendAI Vision One, which pairs hardware-level detection on NVIDIA BlueField DPUs and network security with threat intelligence from the Zero Day Initiative, Trend Micro's vulnerability-research program. The combination is aimed at the moment enterprises hit when agent deployments grow from a handful of pilots into fleets spanning teams and business units — the point at which security becomes the deciding factor in whether agentic AI reaches production at all. TrendAI also frames the move as bringing full AI factory security, protecting the infrastructure where agents are built, trained, and run.

Why Agents Need a New Security Model

An AI agent is made of three parts: the model that reasons, the harness that orchestrates the agent and its sub-agents, and the tools and data the agent can reach. Together they let an agent hold credentials, remember context across sessions, and act on production systems without a human approving every step. That autonomy is exactly what makes agents useful — and risky. TrendAI and NVIDIA agree on the guiding principle: no single control makes an agent secure. Security must be built into every layer the agent depends on, from the silicon up.
The announcement follows NVIDIA's launch of the platform earlier this month, which Paktechstan covered on September 28. For enterprises, the message from both companies is consistent: agents are moving from experiments to infrastructure, and the security has to move with them.
By Hamza Faisal