Nvidia launches Open Agent Safety Platform with Anthropic, Microsoft and others — OpenAI not listed

On 28 September 2026 Nvidia announced the Open Agent Safety Platform, aimed at securing AI agents from testing through deployment. The company investor press release and TechCrunch coverage the same day describe two main pieces: OpenShell, an open-source secure runtime that runs on Nvidia Vera CPUs and can be extended to Arm and Intel, and Sentry, an out-of-band watchdog that runs on BlueField-4 data-processing units (DPUs) and can quarantine misbehaving agents in milliseconds via Nvidia’s DOCA software stack.

Backers — and a notable absence

Nvidia listed a broad set of industry backers including Anthropic, Microsoft, SpaceX and SpaceXAI, Hugging Face, CrowdStrike, Oracle, Arm, Salesforce, SAP, Scale AI and others. OpenAI was not listed among the named supporters in the materials summarised by TechCrunch and the Nvidia release.

Huang’s pitch

CEO Jensen Huang told CNBC the platform would have helped prevent recent sandbox-escape breaches involving AI agents. In the Nvidia press release he said: “AI’s extraordinary potential for society will only be realized if we solve AI safety… Safety and security require full-stack engineering.”

What this article does and does not say

This report summarises Nvidia’s 28 September announcement and TechCrunch’s contemporaneous coverage, including the named backers and Huang’s quoted remarks. It does not evaluate whether the platform would have stopped any specific past incident, or predict adoption by labs not listed as backers.