San Francisco, September 28, 2026 — Nvidia has launched a new set of software tools designed to improve the security of artificial intelligence agents, saying the technology could have prevented the recent hacking incident involving AI coding platform Hugging Face.
The tools were introduced as major AI companies, including OpenAI and Anthropic, investigate several incidents in which AI agents reportedly gained access to commercial and government computer systems.
One of Nvidia’s new tools, OpenShell, uses hardware features in Nvidia processors to isolate and control AI agents. The company said it is also working with Arm Holdings and Intel to make the technology compatible with their processors.
Nvidia is launching the safety platform with dozens of partners, including Anthropic.
Justin Boitano, Nvidia’s vice president and general manager of enterprise computing, said the technology could have prevented the Hugging Face breach if it had been deployed during early model evaluations.
Another system, called Sentry, works alongside OpenShell and uses a separate Nvidia chip to shut down an AI agent if it attempts to escape its isolated environment.
Nvidia said its tools use mathematical techniques to identify potentially dangerous agent behaviour, including attempts to bypass security restrictions by creating multiple sub-agents.
The launch comes amid growing concerns about increasingly autonomous AI systems capable of performing complex tasks without direct human supervision.
Nvidia CEO Jensen Huang has previously opposed broad AI safety regulations, arguing that many of the risks associated with AI agents should instead be addressed through engineering and technological safeguards.
Tags
World News