Palo Alto Networks and NVIDIA want tighter control over AI agents
Palo Alto Networks and NVIDIA are integrating controls that limit what AI agents can access and do.
Palo Alto Networks is expanding its NVIDIA collaboration so organizations can control AI agent activity, network access, and identity. Prisma AIRS AI Gateway on NVIDIA Vera CPUs is described as a future design for inspecting agent interactions, limiting approved tools, and stripping sensitive data. OpenShell is already broadly available and runs agents in isolated environments with file and network policy. Runtime controls on BlueField DPUs are generally available, using DOCA Argus to monitor tool calls and forward telemetry to Cortex XSIAM; an OpenShell integration with Idira for agent secrets is still planned.