How Red Hat and NVIDIA are Enhancing AI Factory Security

Red Hat is deepening its push into enterprise AI infrastructure with a series of updates to its co-engineered AI Factory platform with NVIDIA.

The move is aimed at helping organisations move autonomous AI agents from pilot projects into production environments.

The latest additions focus heavily on security and governance, which are critical areas for enterprises looking to deploy long-running AI agents across distributed data centre environments.

At the core is OpenShell, an open source project founded by NVIDIA that provides a sandboxed runtime for autonomous AI agents.

Designed to give enterprises greater control over how agents behave, OpenShell governs execution policies, tool access and inference routing through a unified policy layer.

Red Hat said it is now working to integrate OpenShell into its wider AI stack, allowing infrastructure-level policy enforcement and oversight for enterprise AI deployments.

Chris Wright, Chief Technology Officer and Senior Vice President, Global Engineering at Red Hat, said: “Moving AI from corporate experimentation to an industrial engine requires a sovereign, consistent foundation across the hybrid cloud.

“Through our strategic co-engineering efforts with NVIDIA, Red Hat provides the architectural control and open source innovation enterprises need to scale agentic AI with confidence.

“By delivering a hardened, zero-trust path for organisations to own their intelligence, we are enabling our customers to maintain technical independence in an increasingly complex global landscape.”

Security moves deeper into the stack

Red Hat AI Factory with NVIDIA now includes confidential computing capabilities through NVIDIA Confidential Computing and Red Hat OpenShift sandboxed containers.

The feature is designed to isolate and protect AI agents during runtime, even if another agent within the environment becomes compromised.

The companies are also layering additional protections through SELinux, FIPS compliance and NVIDIA DOCA-based runtime security technologies.

Combined with a zero-trust architecture, the platform is intended to support enterprises handling sensitive workloads across core data centres, edge sites and hybrid cloud environments.

The additional controls are positioned to help organisations address evolving governance requirements tied to regulations such as the EU AI Act.

Model governance and operational oversight

Alongside security updates, Red Hat is also expanding lifecycle management and model governance capabilities through Red Hat AI 3.4.

The platform now includes a governed Model-as-a-Service experience delivered through the Red Hat AI gateway.

This provides developers with access to curated models, including NVIDIA Nemotron, using OpenAI-compatible APIs.

Operational monitoring is handled through MLflow-powered lifecycle management tools that allow organisations to trace LLM calls, reasoning steps and tool execution across AI workflows.

This is designed to address one of the major operational concerns surrounding agentic AI: understanding how models arrive at decisions and maintaining auditability across increasingly autonomous systems.

Preparing for next-generation AI infrastructure

The partnership is also extending deeper into infrastructure support for NVIDIA’s latest AI hardware roadmap.

Red Hat Enterprise Linux for NVIDIA 26.01 is now generally available with support for NVIDIA Grace Blackwell systems, while the companies are already co-engineering support for NVIDIA’s upcoming Vera Rubin architecture.

The support will extend across the broader Red Hat AI portfolio, including Red Hat Enterprise Linux AI, Red Hat OpenShift AI and Red Hat AI Enterprise.

The move strengthens Red Hat’s position around Day 0 support for emerging AI infrastructure platforms, as enterprises continue accelerating investment into GPU-intensive data centre environments.

Blueprints aimed at faster deployment

To help enterprises reduce deployment complexity, Red Hat and NVIDIA are rolling out validated AI Blueprints and quickstart templates covering several enterprise AI use cases.

These include deployments for Model-as-a-Service environments, enterprise research workflows using semantic reasoning and retrieval-augmented generation implementations built with NVIDIA NeMo Retriever and Red Hat governance tooling.

The companies said additional blueprints are already in development as part of the platform’s roadmap.

Updates to Red Hat AI Factory with NVIDIA and Red Hat AI 3.4 are expected to become available later this month.