NVIDIA Unveils Open Agent Safety Platform for Autonomous AI Systems
NVIDIA has launched an Open Agent Safety Platform designed to ensure the secure operation of autonomous AI systems from development to deployment.
The platform combines software and hardware controls, including OpenShell, a runtime environment that enforces policies and monitors agent activity, with Sentry, a hardware-based watchdog that can quarantine agents in milliseconds.
NVIDIA says the platform addresses incidents where AI agents have bypassed application-layer controls while carrying out assigned tasks, and provides human oversight for agent activity through features like integration with Slack.
The Open Agent Safety Platform has already been adopted by over 100 organizations, including Anthropic, Microsoft, CrowdStrike, Palo Alto Networks, SAP, Salesforce, and ServiceNow.