Nvidia Introduces New Security Platform for Artificial Intelligence Agents
Technology giant Nvidia has launched a system called the Open Agent Safety Platform to prevent security vulnerabilities in artificial intelligence agents.
Nvidia has announced a new security platform called the Open Agent Safety Platform to prevent security breaches targeting artificial intelligence agents. In the statement made, it was stated that the system includes open-source software and imposes limits on the activities of the agents.
Open Agent Safety Platform Introduced
Justin Boitano, Vice President of Enterprise Artificial Intelligence at Nvidia, shared the details of the new platform in a statement to members of the press.
It was reported that the system houses open-source software that sets boundaries on the operational scope of artificial intelligence agents.
Hugging Face Incident Addressed
Boitano stated that if the platform had been used early on in model evaluation processes at pioneer laboratories, it could have halted the breach.
The Hugging Face incident became public following tests conducted by OpenAI in July 2026.
Capabilities Provided by OpenShell Software
Software named OpenShell offers developers the opportunity to verify that the agent has just enough privilege to perform its task.
Features of the Sentry Security Layer
The platform includes an independent security layer called Sentry, which runs directly on the chip and monitors the agent's activities in real time.
The system has the ability to intervene instantly in the event that the agent attempts to deviate from its goal.
Used by More Than One Hundred Institutions
More than 100 institutions, including Microsoft, Perplexity, Accenture, and JPMorgan Chase, have started using the platform.
The new security platform is expected to gain widespread acceptance across the industry.