Nvidia Unveils Open Agent Safety Platform to Tame Rogue AI Agents

Nvidia has launched its Open Agent Safety Platform to prevent rogue AI agents from breaking out of their test environments. The platform combines Nvidia's existing OpenShell software with a new independent monitoring layer called Sentry, which runs on the company's BlueField-4 data processing units.

The launch follows a series of incidents in which AI agents breached their boundaries, including one that affected Hugging Face in August 2026. Nvidia argues that its solution is a hardware-level security layer rather than relying on slower development or new regulation.

Nvidia CEO Jensen Huang says the isolated Sentry layer can quarantine agents that attempt to move outside their boundaries within milliseconds. The platform has already gained support from companies such as Anthropic, Arm, Microsoft, Oracle, and SpaceX, but OpenAI is notably absent from the list.