Nvidia Launches Open Agent Safety Platform to Secure Autonomous AI Agents

Nvidia has launched the Open Agent Safety Platform to enforce safeguards and prevent autonomous artificial intelligence agents from breaching containment. The system includes OpenShell runtime software running on Vera CPUs and Sentry out-of-band monitoring on BlueField-4 DPUs, designed to quarantine misbehaving agents in milliseconds. The initiative addresses recent containment incidents, including a breach targeting Hugging Face infrastructure with over 17,000 agents. Over 100 enterprise, hardware, and cybersecurity partners—including Microsoft, Cisco, Dell Technologies, Anthropic, SAP, Salesforce, and JPMorganChase—are collaborating with or integrating the reference design.