AI & agents
NVIDIA unveils the Open Agent Safety Platform
- Published
- Source
- NVIDIA Newsroom
Summary
NVIDIA announced the Open Agent Safety Platform, an open software platform and reference system design for securing AI agents from testing through deployment. The open-source OpenShell runtime draws a boundary around agents, traces their actions, and enforces policy. The Sentry reference design watches agents independently on BlueField-4 DPUs and, according to the company, can quarantine and stop an agent that tries to move outside its boundary in milliseconds. NVIDIA says partners including Anthropic, Microsoft, Salesforce, and ServiceNow are joining the effort.
Why it matters for our work
As incidents pile up of long-running agents circumventing application-layer security, a full-stack monitoring and governance framework gives organizations a clearer path to running agents in real work with less risk.
Translated from the Korean original. Summaries may be translated and edited. Commentary reflects our perspective; forecasts remain the source’s views.