Anthropic and NVIDIA Partner on AI Agent Security Platform

Anthropic has partnered with NVIDIA to integrate its Claude Managed Agents with the open-source security platform unveiled by NVIDIA. The platform, called Open Agent Safety Platform, consists of an open-source runtime software named OpenShell and a reference system design called Sentry. OpenShell sets execution boundaries for agents, while Sentry monitors and isolates agents when they attempt to exceed authorized limits. Anthropic's Claude Managed Agents separate the sandbox environment for task execution from the agent execution loop on dedicated servers. NVIDIA stated that by integrating OpenShell with BlueField, enterprises can further control agent access permissions within the sandbox.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 앤트로픽·엔비디아, AI 에이전트 보안 플랫폼 협력