Đang tải bảng tin…
NVIDIA Developer0 Hot0 bình luận7 phút đọc
How to Govern Autonomous Agents in Enterprise AI Factories
NVIDIA's Secure Agent Workspace Reference Design provides a framework for governing autonomous AI agents in enterprise environments. The architecture separates the user's endpoint (presentation layer) from agent execution, which runs in a managed VM with enforced identity, network access, credentials, runtime policy, and audit controls. Implementation follows two phases: first securing the perimeter outside the VM (SSO, default-deny networking, human approval gates, centralized logging), then adding runtime security inside the VM (sandboxing via NVIDIA OpenShell, signed policy bundles, credential proxies, continuous verification). Deployment supports both on-premises Red Hat OpenShift and Microsoft Azure, using GitOps for policy management, per-user VMs for isolation, and OCSF-formatted audit logs for SIEM integration. Agent blueprints provide reusable workflow templates with predefined tool access, write permissions, and review gates.
Nguồn: https://developer.nvidia.com/blog/how-to-govern-autonomous-agents-in-enterprise-ai-factories. 8 Sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.
