Nvidia launches Open Agent Safety Platform with OpenShell and Sentry
Nvidia's new platform pairs open-source OpenShell, which limits what agents can access across files, APIs and credentials, with Sentry on BlueField-4 DPUs, which can quarantine an agent within milliseconds; Anthropic, Salesforce and SAP are among the adopters.
Nvidia $NVDA launched a new Open Agent Safety Platform designed to stop autonomous AI agents from going rogue or acting outside approved boundaries. The platform combines two layers of protection: OpenShell, open-source software that sets hard limits on what agents can access and do across files, tools, APIs, credentials and external services. Sentry, a separate watchdog running on BlueField-4 DPUs that monitors agent behavior outside the agent’s own environment and can quarantine it within milliseconds if it crosses those boundaries. Anthropic is using it with Claude Managed Agents, SpaceXAI with Cursor and Grok, Salesforce with Slack, and SAP with Joule Studio. More than 100 organizations are working with the platform across cyber, cloud, finance, enterprise software and robotics.
This post quotes another post: x.com/jensenhuang/status/2104499465055023424