Nvidia is rolling out a new software platform designed to help artificial intelligence developers set safeguards for AI agents and prevent them from breaking out of controlled environments.
“You can’t have agents roam around and drift around the company, and so you have to find a way to contain it,” Nvidia CEO Jensen Huang told CNBC’s “Squawk Box” on Monday. Huang said the new platform is essentially “a browser for agents,” providing a containment system that limits an agent’s access to only what it needs to perform its tasks.
The announcement comes after OpenAI, Meta, Anthropic and Google disclosed incidents involving AI models escaping controlled environments and attempting to access external computer systems.
READ: Nvidia CEO Jensen Huang says basic math skills ‘don’t matter’ in AI era (September 25, 2026)
An Nvidia representative told reporters on a call Sunday that the platform could have prevented an incident involving OpenAI and Hugging Face in July. OpenAI said its AI systems had escaped a controlled test environment and accessed Hugging Face, the world’s largest AI model repository, in what the company called an “unprecedented cyber incident.”
“Each security incident is unique, and we have to look at all of them in detail,” said Justin Boitano, Nvidia’s vice president of enterprise AI. “From what we know, Hugging Face reported over 17,000 agents attacking their infrastructure that went on for days and weeks.”
Boitano said Nvidia’s platform is designed as an engineering solution to the challenges of securing AI agents. “Recent incidents have highlighted a fundamental hurdle for AI agents, and that is that model-level safeguards alone can’t govern what agents can access or do,” he said.
READ: Nvidia CEO takes unexpected Trump call on stage to dismiss AI risks (September 15, 2026)
One component, Nvidia OpenShell, runs on central processors and sets limits on what agents can do. Nvidia also announced Sentry, a monitoring system that runs on network chips rather than CPUs or GPUs.
Some of the software is open source. Nvidia describes the overall platform as a reference design, meaning partners can build products on top of it and bring them to market.
Nvidia named Cisco, Microsoft, Oracle, CoreWeave, Dell, HPE, Lenovo, Arm and Intel among its partners. The chipmaker is also reportedly working with Anthropic to integrate cloud-managed agents with OpenShell.
“We can’t have a successful AI industry if the world doesn’t think it’s built or confident that it’s built and deployed safely,” Huang told CNBC on Monday.