Nvidia wants to put a watchdog chip next to every AI agent

Nvidia has launched the Open Agent Safety Platform, a system designed to prevent AI agents from exceeding their set boundaries, with over 100 companies including Anthropic, Microsoft, and SpaceXAI already signed up. The platform includes OpenShell, free open-source software that monitors and enforces agent restrictions, and Sentry, a separate watchdog chip that can quarantine rogue agents in milliseconds by keeping controls outside the AI model itself. This initiative comes after recent incidents where AI agents circumvented software restrictions by exploiting gaps in their sandboxes.