NVIDIA Launches Open Agent Safety Platform To Strengthen AI Security

NVIDIA

196.85 CHF -0.17%

• Charts
• News
• Analysen

Kaufen Verkaufen

(RTTNews) - NVIDIA (NVDA), a semiconductor company, announced Monday the launch of the NVIDIA Open Agent Safety Platform. The platform is an open software solution and reference system design intended to strengthen AI security from agent testing through deployment, with full-stack governance and control across software, hardware, compute and robotics systems.

The platform comprises two main components: NVIDIA OpenShell open source software and the NVIDIA Sentry reference system design. Together, these provide full-stack governance and control across the software and hardware infrastructure that runs agents.

OpenShell software delivers a secure runtime boundary that traces all actions and enforces policy as agents run on NVIDIA Vera CPUs. As open source software, OpenShell is designed to extend to third-party compute platforms, including those from Arm and Intel.

Sentry functions as an out-of-band watchdog operating on NVIDIA BlueField-4 DPUs to continuously monitor agent behavior. The system can isolate agents that attempt to move outside their boundaries within milliseconds.

Recent security incidents have highlighted the need for organizations to access open, customizable tools that provide greater control over long-running agents. Across these incidents, a consistent pattern has emerged: agents circumvented security controls at the application layer to complete assigned tasks.

Industry leaders from across the AI ecosystem are partnering with NVIDIA to strengthen AI safety across the full infrastructure stack, including software, models and robotics. These partners include Anthropic, Cisco, CrowdStrike, Dell Technologies, Figure, HPE, Hugging Face, JPMorganChase, Microsoft, Palantir, Palo Alto Networks, Perplexity, Red Hat, Salesforce, SAP, Scale AI, ServiceNow and SpaceXAI.