# Red Hat and Nvidia expand partnership to secure autonomous AI agents

> Red Hat added new runtime and hardware controls to Nvidia’s platform, aiming to stop AI agents from overstepping their limits or being hijacked.

Oossa · 2026-10-21 · https://oossa.com/en/red-hat-and-nvidia-expand-partnership-to-secure-autonomous-ai-agents

Red Hat announced at Nvidia’s GTC event in Berlin that it will integrate Nvidia’s OpenShell sandbox and Sentry security framework into its AI Factory platform. The joint design places protection outside the agent, using Nvidia’s BlueField data processing units to monitor memory and network activity without host‑level software. The companies also added support for Nvidia’s Nemotron 3.5 Super VL model in the vLLM inference engine and prepared Red Hat Enterprise Linux for the upcoming Vera Rubin data‑center hardware.

## The facts

- Red Hat OpenShift 4.22 already supports BlueField‑3 and the new Sentry can quarantine a compromised agent in milliseconds.
- Joint reference design adds Model‑as‑a‑Service controls and supports Nvidia Nemotron 3.5 Super VL in vLLM.

## Why it matters

Enterprises get a hardware‑level guard that can quickly isolate rogue AI agents, reducing the risk of data breaches or misuse.

## Sources & references

1. [Red Hat and Nvidia expand their partnership to secure autonomous AI agents](https://siliconangle.com/2026/10/21/red-hat-and-nvidia-expand-their-partnership-to-secure-autonomous-ai-agents/) – SiliconANGLE, 2026-10-21

Last updated: 2026-10-21