Vera BlueField-4 STX adds DOCA Vault, Argus and Flow to secure agentic AI storage

At GTC Taipei, NVIDIA introduced three new DOCA microservices for its Vera BlueField-4 STX platform — DOCA Vault, DOCA Argus and DOCA Flow-positioning storage as a primary control point for enterprises deploying agentic AI that reason over, retrieve and modify business context. The company says these services are built to enforce policies and inspections inline on the storage datapath in BlueField-4 silicon, so enterprises can govern autonomous agents without slowing AI data flows.

DOCA Vault is designed to ensure that only authorized AI workloads access files with the correct permissions. DOCA Argus provides visibility into agent behavior and workload activity to detect anomalous or risky actions. DOCA Flow isolates network traffic and protects sensitive data in multi — tenant environments by enforcing network — level policies between agents and the data they touch.

NVIDIA says running policy enforcement and inspection in BlueField-4 silicon enables runtime threat detection it claims can be up to 1,000x faster than existing agentless runtime solutions, and can enforce network and file access policies at speeds up to 800 Gb/s. The release responds to a shift from simple chatbots to multi — step autonomous agents that continuously read, write and share proprietary context memory — creating new exposure points that storage — level controls aim to close.

The announcement is presented as part of an ecosystem strategy: cybersecurity integrators working 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. Storage vendors building STX-based platforms include Cloudian, DDN, Dell Technologies, Everpure, Hitachi Vantara, HPE, IBM, MinIO, NetApp, Nutanix, VAST Data and WEKA. Hardware manufacturers and system builders listed are AIC, ASUS, Foxconn, Gigabyte, QCT, Supermicro, Wistron and Wiwynn, with systems integrators such as Accenture and Deloitte also involved.

For builders and platform teams, NVIDIA frames the update as a way to enforce DOCA libraries and microservices in silicon while data continues to move at AI factory speeds — allowing inline inspection and governance between agents, data and long-context memory. That approach is presented as reducing the risk of unauthorized access, data extraction and sensitive — context exposure without adding operational complexity to enterprise AI deployments.