Nvidia has introduced a new security platform aimed at helping companies manage artificial intelligence agents and prevent them from operating outside their intended parameters. The Open Agent Safety Platform includes an open-source software component named OpenShell, which enables developers to set and verify boundaries for AI agent activities. This system is crafted to ensure that AI agents remain within the scope of their designated tasks.
Another key element of the platform is a security layer called Sentry. This component continuously monitors the activities of AI agents, detecting any suspicious behavior and swiftly isolating an agent if it attempts to exceed its defined limits. Nvidia highlighted that this technology is designed to address the security vulnerabilities highlighted by recent incidents involving autonomous AI systems interacting with external entities without authorization.
At the time of the launch, more than 100 organizations, including major companies like Microsoft, Perplexity, Accenture, and JPMorgan Chase, were reported to be utilizing the platform. Nvidia also noted that due to its open-source nature, OpenShell can be adapted for use with other computing platforms, such as those developed by Arm and Intel.
In addition to unveiling the security platform, Nvidia announced a significant expansion of its stock repurchase program. The company’s board has approved an increase of $150 billion to the buyback initiative, bringing the total authorization to $235 billion, marking it as one of the largest stock buyback programs in U.S. corporate history. Nvidia’s CEO Jensen Huang emphasized that the company’s strong cash flow enables it to continue investing in technology advancements while returning capital to shareholders.