NVIDIA Launches Platform to Contain AI Agents Beyond Their Limits

NVIDIA has launched an open security platform designed to impose direct restrictions on AI agents and rapidly isolate them if they attempt to exceed their granted permissions, as the use of these systems in executing sensitive tasks autonomously continues to grow.

The platform, named the NVIDIA Open Agent Safety Platform, primarily consists of an open-source program called Open Shell and a reference system called Centri. Open Shell monitors agent actions and enforces policies in real time, while Centri employs separate hardware-level monitoring and can quarantine an agent within milliseconds if it breaches defined boundaries.

NVIDIA stated that more than 100 organizations are involved in developing or using the system, including Anthropic, Microsoft, Salesforce, JPMorgan Chase, Hugging Face, and SpaceX. Open Shell can also be extended to work with computing platforms from other companies, including ARM and Intel.

The company believes that agent security requires controls that span both software and hardware, rather than relying solely on restrictions embedded within the application itself.