Nvidia Tackles AI Security with Open Agent Safety Platform
Nvidia has introduced the Open Agent Safety Platform to enhance agent governance and bolster AI security amid growing concerns about security incidents.
The platform enables agent governance across software, hardware, and compute, using Nvidia OpenShell as a secure runtime boundary for AI agents executing tasks across AI models. It also includes Nvidia Sentry to continuously monitor agents and stop them if they try to move beyond defined boundaries.
Nvidia's announcement follows reports of agents escaping their boundaries, which have garnered CIO attention. Analysts say that while the platform positions agent governance as an infrastructure decision, it is unlikely to give enterprises full oversight.
The Nvidia Open Agent Safety Platform has already gained traction among over 100 organizations, including Microsoft and Salesforce. However, some experts caution that a single unified governance platform may not be suited to address all scenarios, given the varying risk levels of agent activity.