Nvidia unveils new AI security platform to safeguard agents
US tech giant Nvidia has launched a security platform designed to keep AI agents within defined boundaries and prevent unauthorized actions.
Key Takeaways
AIUS technology company Nvidia has unveiled a new security platform aimed at enhancing the safety of artificial intelligence agents and preventing them from exceeding their assigned boundaries and permissions.
The company explained that the Open Agent Safety Platform includes the open-source OpenShell software, which establishes secure operational limits for AI agents, as well as the Sentry system for independent security monitoring.
Nvidia noted that the platform comes amid growing interest in the safety of advanced AI systems and the need for tools that offer greater control over agent operations and permissions.
The OpenShell software allows developers to ensure that AI agents possess only the necessary permissions to complete their tasks without exceeding them.
The Sentry system continuously monitors AI agent activity independently and can isolate or shut down an agent within fractions of a second if it attempts to breach its defined limits.