NVIDIA adds security features to Vera BlueField-4 STX
Tue, 2nd Jun 2026 (Today)
NVIDIA has introduced new DOCA security features for its Vera BlueField-4 STX platform, aimed at storage systems used for agentic AI.
The update focuses on software- and silicon-based controls designed to manage how autonomous AI agents access files, interact with data and move across networks. New and enhanced functions include DOCA Vault for file access controls, DOCA Argus for visibility into agent and workload activity, and DOCA Flow for network isolation in shared AI environments.
NVIDIA is positioning Vera BlueField-4 STX as a storage processing platform for enterprises building AI systems that do more than answer prompts. In this model, agents retrieve information, write data and take actions across corporate systems, creating new security concerns around context memory, file access and continuous machine-driven activity.
The controls are enforced directly in BlueField-4 silicon rather than handled only by external software layers. NVIDIA said this is intended to allow inspection and policy enforcement in the data path while storage traffic continues to move at high speed.
According to NVIDIA, Vera BlueField-4 STX can deliver runtime threat detection up to 1,000 times faster than existing agentless runtime tools. The platform can also enforce file and network access policies at speeds of up to 800Gb/s.
Security focus
The announcement reflects a broader shift in enterprise AI from chatbot deployments to systems in which software agents operate with greater autonomy. These systems depend on rapid access to proprietary data and stored context, but they also raise the risk of unauthorised access, data extraction and exposure of sensitive information if controls are weak.
By placing policy enforcement in the storage and networking layer, NVIDIA is seeking to make security a built-in part of AI infrastructure rather than an added monitoring tool. The expanded Vera BlueField-4 STX platform brings DOCA security libraries and microservices into the AI storage layer to help enterprises protect data, agents and context memory as these systems move into production.
"Agentic AI turns enterprise data into a living, real-time system - and that system must be protected where data moves, where context is stored and where agents act," said Jensen Huang, founder and chief executive officer of NVIDIA.
"With Vera BlueField-4 STX, NVIDIA and its ecosystem are building secure-by-design storage infrastructure that enforces trust in silicon at the speed of AI," Huang said.
Partner network
NVIDIA also used the launch to highlight a broad partner group around the platform. Cybersecurity companies integrating enterprise security tools with Vera BlueField-4 STX include Akamai, Armis from ServiceNow, Check Point, Cisco, CrowdStrike, EQTY, F5, Fortinet, Palo Alto Networks, TrendAI, Xage Security and Zscaler.
On the storage side, Cloudian, DDN, Dell Technologies, Everpure, Hitachi Vantara, HPE, IBM, MinIO, NetApp, Nutanix, VAST Data and WEKA are building platforms based on STX. NVIDIA said these systems are aimed at inference, training and analytics workloads.
Hardware manufacturing partners include AIC, ASUS, Foxconn, Gigabyte, Quanta Cloud Technology, Supermicro, Wistron and Wiwynn. Systems integrators Accenture, Deloitte and Worldwide Technology are also working with NVIDIA and its partners on enterprise deployments.
Infrastructure push
The move adds to NVIDIA's effort to extend its reach beyond AI chips into the broader infrastructure stack for enterprise AI. Storage has become a more strategic part of that stack as organisations try to keep large volumes of data close to models and agents that need fast retrieval and longer context windows.
For customers, the practical question is whether silicon-level enforcement can reduce operational overhead and improve control without slowing workloads or adding another layer of management complexity. NVIDIA argues that combining storage acceleration with embedded security is intended to address both performance and governance demands as agentic AI deployments grow.
Partners are building enterprise-scale platforms around Vera BlueField-4 STX, with systems expected to become available in the second half of 2026.