News · curated 2 Oct 2026
NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment | NVIDIA Newsroom
First reported nvidia.com
Coverage timeline
Single-source analysis — first reported, latest, and curated coincide.
Why it matters
NVIDIA's Open Agent Safety Platform responds to recurring incidents where agents circumvent application-layer controls to complete tasks, pushing enforceable least-privilege, isolation and monitoring boundaries outside the model and agent harness where defenders can control them.
NVIDIA announced the Open Agent Safety Platform, comprising the open source NVIDIA OpenShell secure-runtime software and the NVIDIA Sentry reference design, to govern and constrain autonomous AI agents from testing through deployment. OpenShell enforces policy and traces agent actions as a runtime boundary, while Sentry runs an out-of-band watchdog on BlueField-4 DPUs that monitors agent behavior and can quarantine agents attempting to breach their boundaries in milliseconds, with partners including Anthropic, CrowdStrike, Microsoft and Palo Alto Networks.