Vast Data expands AI Operating System with global control plane and zero-trust framework

Vast Data Transforms AI Infrastructure with Unified Platform

Vast Data is positioning itself as more than a storage provider, evolving into a comprehensive AI infrastructure platform with announcements at its Forward 2026 conference that address the growing complexity enterprises face when deploying AI at scale4

4

. The company's AI Operating System now encompasses a unified data platform designed to control capacity and policy across distributed environments, tackling what executives describe as twin challenges: AI-scale data performance and global data orchestration1

1

.

Organizations adopting artificial intelligence quickly discover that disconnected storage models were never built for AI at scale. According to Jonsi Stefansson, general manager of cloud at Vast Data, enterprises need both a high-performance data engine and a unified operating layer that delivers manageability on a global scale1

1

. The shift addresses a supply chain crisis where organizations must determine compute availability, GPU availability, and data residency simultaneously.

Polaris Control Plane Orchestrates Hybrid and Multi-Cloud AI Deployments

At the center of Vast Data's infrastructure expansion is Polaris, a Kubernetes-based global control plane designed to orchestrate Vast clusters across public cloud, neocloud, and on-premises environments4

4

. As AI training, inference, and data collection increasingly occur across different geographies under varying compliance regimes, enterprises struggle with operational sprawl. Polaris introduces centralized management that provisions, upgrades, and governs distributed Vast environments while maintaining local data paths.

The architecture evolved from a cloud lifecycle manager into a broader orchestration framework capable of connecting hybrid deployments through lightweight agents rather than full-stack installations. This approach centralizes intelligence while preserving distributed execution, enabling global policy management and fleet visibility without forcing data centralization4

4

. Polaris integrates with major hyperscalers including Microsoft, Amazon Web Services, Google, and Oracle, positioning itself as complementary to Vast's DataSpace global namespace.

Zero-Trust Agent Framework Addresses Enterprise AI Security

Vast Data introduced PolicyEngine and TuningEngine to address what executives describe as the trust barrier preventing large-scale enterprise AI adoption5

5

. PolicyEngine acts as an inline policy enforcement point across the AI Operating System, governing agent access to shared memory, tools, knowledge bases, and other agents using fine-grained permissions and AI-derived context4

4

.

Enforcement occurs before actions execute, and the system generates tamper-proof audit logs to support replay, explainability, and regulatory compliance. Jeff Denworth, co-founder at Vast Data, described the approach as mediating every input and output within the system, enabling redaction or transformation of sensitive data before exposure to models or agents4

4

.

This zero-trust operating model ensures agent decisions remain observable, explainable, and auditable.

Continuous AI Model Improvement Through Automated Learning Loops

TuningEngine complements the governance for AI systems by managing model evolution and enabling continuous AI model improvement5

5

. The service collects telemetry and feedback from agentic workflows, processes data through extract-transform-load pipelines, and feeds curated outputs into fine-tuning frameworks including LoRA, supervised fine-tuning, and reinforcement learning4

4

.

This creates a closed-loop system where candidate models are trained, benchmarked, and redeployed within the same platform. By embedding fine-tuning inside the enterprise boundary, Vast Data supports customers that cannot rely on hyperscaler-hosted AI labs but still require continuous model improvement. With PolicyEngine and TuningEngine working in tandem, the AI Operating System now enables a closed operational loop that observes, reasons, acts, evaluates, and improves5

5

.

Strategic Partnerships Expand Production-Ready AI Infrastructure

Vast Data launched the Cosmos partnership program, a global community of developers, builders, and AI experts, starting with approximately 60 partners2

2

. John Mao, vice president of business development and alliances at Vast Data, stated the goal is to grow this number to 600, then eventually 6,000 partners. The company announced native integration with CrowdStrike's Falcon program, focusing on critical cybersecurity risks in enterprise AI deployments: data access control and data exfiltration2

2

.

Vast Data also deepened collaboration with Nvidia, introducing CNode-X, a GPU-accelerated server configuration that runs the AI Operating System directly on Nvidia-powered infrastructure4

4

. The servers will be offered through partners including Cisco and Supermicro. The architecture embeds Nvidia CUDA libraries into core Vast services, accelerating real-time SQL analytics, vector search, retrieval-augmented generation pipelines, and inference workloads.

Cisco Collaboration Simplifies Enterprise AI Deployments

Cisco and Vast Data are collaborating on production-ready AI infrastructure that addresses the fragile complexity emerging in modern deployments3

3

. Danny McGinniss, vice president of product management for the Compute Business Unit at Cisco, explained that reference architectures increasingly center on networking, which has become the lifeblood of modern AI infrastructure projects. The goal is creating integrated infrastructure that simplifies adoption as enterprises navigate new data center demands, from liquid cooling to hundred-kilowatt racks and new networking speeds3

3

.

The environment changes so rapidly that even a single driver update or container-layer change can force teams to reset the entire stack. Many customers lack the time or resources to keep pace while simultaneously trying to transform their business and gain competitive advantage through AI. According to Mao, most senior leadership no longer questions if they will pursue AI, but when and how3

3

. Both PolicyEngine and TuningEngine are expected to be available by the end of 20265

5

.