TrendAI Vision One ™ will combine hardware-level detection and network security based on NVIDIA BlueField DPU with Zero Day Initiative ™ threat intelligence to help enterprises safely advance their proxy clusters from pilot projects to production use.
Dallas, September 28, 2026 / PRNewswire / -- TrendAI ™ is a global AI security leader and business unit under Trend Micro Incorporated (Tokyo Stock Exchange Code: 4704; TSE : 4704). The company today announced its support for NVIDIA Agent Safety Platform, which is a full-stack infrastructure for enterprises to operate autonomous AI proxies. As organizations put autonomous, self-evolving proxies into production, TrendAI ™ aligns with NVIDIA on one key principle: no single control measure can make a proxy secure. Security must be embedded at every layer that the proxy relies on.
AI Agents consist of three core components: a model responsible for reasoning, a framework that orchestrates the agents and their sub-agents, and the tools and data that the agents can access. Together, these components enable agents to hold credentials, remember context across sessions, and take action on production systems without the need for manual step-by-step approval. As enterprises expand from small pilots to groups of agents across teams and business departments, security has become a key control factor and often determines whether agent-based AI can be deployed in a production environment.
NVIDIA Agent Safety Platform places permissions within the infrastructure beneath the proxy. According to NVIDIA, OpenShell executes policies outside of the proxy execution environment, hence it is not possible to bypass or evade these policies through prompts; NVIDIA BlueField -4, DPU, and NVIDIA DOCA add an additional security layer at the silicon level that is independent of the host. TrendAI ™ serves to supplement this execution layer by providing intelligence to determine what policies should be implemented, offering visibility to observe what the proxy is doing, and providing threat detection and response capabilities to take action.
TrendAI ™ Head of Platform and Business Officer Rachel Jin stated: "Proxy AI represents the most important transformation in enterprise technology over a generation, and security requirements and safeguards must be in place to ensure that proxies align with organizational intentions. Security cannot be focused in just one area; it must cover models, frameworks, as well as every tool and data source that proxies come into contact with. NVIDIA is working to establish the boundaries that determine what proxies can do. TrendAI ™ provides threat intelligence and monitoring to inform the security team about what these boundaries should be and what is happening within them. Together, we offer an actionable path for organizations to safely scale proxy AI."
How can TrendAI support NVIDIA Agent Safety Platform?
TrendAI Vision One ™ extends the proxy protection of NVIDIA to the entire AI factory, covering proxies, models, containers, storage, and networking on a single platform:
- Model protection. Inline checks on prompts and responses can prevent prompt injection and jailbreak attacks, and mask sensitive data before it is leaked.
- Framework and tool governance. The security team can control which tools and MCP servers proxies have access to, inspect each tool invocation, and analyze the behavior of skills and AI workpieces during runtime.
- Data and identity control. TrendAI Vision One ™ can identify and classify sensitive data provided by AI, prevent proxies from accessing data they should not have access to, extend file security and data leakage prevention to files read and written by proxies, and remove excessive permissions granted to proxies and other non-human identities.
- Hardware-based detection and response. With the help of NVIDIA BlueField DPU and DOCA telemetry, even if the host operating system is compromised, the detection provided by TrendAI Vision One ™ can remain reliable. It also associates AI signals with terminal, network, and identity telemetry, preventing hijacked proxies from moving laterally without being detected.
TrendAI ™ and NVIDIA jointly lead the future of AI with proactive security, aiming to stimulate innovation and eliminate risks.
Learn more
If you want to understand why proxy security must go beyond the model itself, you can read the blog post Rachel Jin titled "Beyond Model Alignment : Securing Every Layer of the AI Agent". Here, you can learn how TrendAI Vision One ™ protects AI.
About TrendAI ™
TrendAI™ is a global AI security leader and enterprise business unit under Trend Micro. It helps organizations achieve comprehensive AI visibility and integrated security capabilities to build confidence, drive innovation, and eliminate risks. TrendAI™ is trusted by the largest enterprises and government agencies in 185 countries, protecting entire organizations from identity to infrastructure to data. Fortune Global 500 companies rely on TrendAI™ to reduce risks and, with world-leading threat and attack intelligence, can prevent threats up to 3 months in advance.
Through deep ecological cooperation with market leaders such as NVIDIA, Anthropic, AWS, Google, and Microsoft, TrendAI™ helps your organization move forward safely at the speed of AI. For more information, please visit trendaisecurity.com.
