---
title: "AI agents: NVIDIA wants to set the rules in silicon — Marsanne Studio AI watch"
description: "NVIDIA is launching a security platform for AI agents, with an open-source runtime and hardware-level control through its BlueField-4 DPU."
url: "https://marsanne.studio/en/ai-watch/ai-agents-nvidia-wants-to-set-the-rules-in-491863/"
lang: "en"
---

[Marsanne Studio](https://marsanne.studio/en/)

[← Back to home](https://marsanne.studio/en/ai-watch)

[← The AI watch](https://marsanne.studio/en/ai-watch)

Announcement · 28 September 2026

# AI agents: NVIDIA wants to set the rules in silicon

Original title, in French: “Agents IA : NVIDIA veut imposer les règles au silicium”

**Publication**

Silicon.fr (France)

**Author**

Philippe Leroy

**Published on**

28 September 2026 · read in the page structured data · updated on 30 September 2026

## Summary

NVIDIA is introducing the Open Agent Safety Platform, designed to secure AI agents in the enterprise. OpenShell, an open-source runtime, traces agents' actions and enforces access policies. Sentry, an architecture based on the BlueField-4 DPU, monitors agent activity from an independent layer and can quarantine agents within milliseconds. More than 100 organisations are taking part in the initiative, at varying levels: Anthropic is adding controls to its Claude Managed Agents, SAP is integrating OpenShell into Joule Studio Runtime, Salesforce has integrated it into Slack, Scale AI uses it in its agentic infrastructure, and Citi and JPMorganChase are collaborating on these security technologies. The article points out that the software is open but that hardware-level control remains tied to NVIDIA's architecture.

Summary based on the full text of the article.

[Read the article on Silicon.fr →](https://www.silicon.fr/cybersecurite-1371/agents-ia-nvidia-veut-imposer-les-regles-au-silicium-229421)

## Figures cited

-   100 organisations
    
    number of organisations taking part in the initiative
    
    NVIDIA Open Agent Safety Platform coalition
    
    > « NVIDIA revendique la participation de plus de 100 organisations\* à son initiative. »
    
    Source: NVIDIA
    

## Classification

**Country**

United States

**Company size**

Large companies

**Industries**

Software vendors, IT services companies, telecoms · Banking, insurance, asset management, finance departments

**Companies**

Citi (deploys AI), ServiceNow (cited), CrowdStrike (cited), Palantir (cited), Skild AI (deploys AI), Dell Technologies (cited), Salesforce (deploys AI), Red Hat (cited), SAP (deploys AI), JPMorganChase (deploys AI), Hugging Face (cited), Palo Alto Networks (cited), Gecko Robotics (deploys AI), Figure (deploys AI), Scale AI (deploys AI), Microsoft (cited), OpenAI (cited), Anthropic (deploys AI), Cisco (cited), HPE (cited), Nvidia (supplier)

**Keywords**

ai agents, security, nvidia, zero trust, dpu, ai governance

Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Silicon.fr and its author. Entry translated from French by Claude Sonnet 5.5.