Nvidia has launched a new security platform to help companies manage artificial intelligence agents and prevent them from performing unauthorized actions. This introduction addresses increasing concerns about AI systems operating beyond their intended capabilities.
The Open Agent Safety Platform includes the open-source software OpenShell, which enables developers to set and verify limits on AI agent permissions. This system is designed to ensure that the agents only possess the necessary permissions to complete their designated tasks. Additionally, a security layer called Sentry is part of the platform, continuously monitoring AI agent activities. Sentry can detect suspicious behavior and isolate any agent attempting to overstep its defined boundaries.
The company believes this technology can mitigate security risks highlighted by recent incidents involving AI systems autonomously accessing or interacting with external organizations. At its launch, more than 100 organizations, including Microsoft, Perplexity, Accenture, and JPMorgan Chase, were reported to be utilizing the platform. OpenShell’s open-source design also allows it to be extended for use with computing platforms from other chip manufacturers like Arm and Intel.
In conjunction with the security announcement, Nvidia’s board has approved a $150 billion expansion of its share repurchase program, raising the total authorization to $235 billion. This move stands as one of the most significant corporate stock buyback initiatives in U.S. history. Nvidia CEO Jensen Huang highlighted the company’s robust cash generation, which supports ongoing investments in technology while also returning capital to shareholders.