New NVIDIA Open Agent Platform for AI Safety

NVIDIA has announced the NVIDIA Open Agent Safety Platform, an open software platform and reference system design aimed at enhancing security from the testing to the deployment of AI agents.

New Platform for AI Safety

NVIDIA has announced the NVIDIA Open Agent Safety Platform to ensure and monitor the security of AI agents as they undertake autonomous tasks. This platform aims to provide a comprehensive security infrastructure from the testing phase to deployment.

Identity and Access Management Integrations

IBM Agent Identity and HashiCorp Vault are integrated with NVIDIA OpenShell to ensure that every agent is assigned a verified identity and granted only the access it requires. IBM Identity Protection also helps monitor all agents, including confidential agents.

Infrastructure and Hardware-Level Protection

IBM Storage integrates NVIDIA BlueField-4 at the hardware level, creating an additional layer for data accessed by agents. Red Hat OpenShift provides the necessary infrastructure to run monitored agents in a hybrid cloud environment.

Zero Trust and Rapid Response Systems

NVIDIA Sentry uses NVIDIA DOCA on BlueField-4 for authentication, enforcing access policies, and providing telemetry. Sentry can quarantine compromised agents within milliseconds.