Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’
Nvidia's new safety platform claims to quarantine rogue agents in milliseconds—but the real test is whether enterprises can trust it at scale.
Why it matters
Nvidia launches Open Agent Safety Platform with OpenShell and Sentry technology to contain and monitor AI agents in real time. The move directly addresses escalating agent-escape incidents and adds boundary enforcement infrastructure to production deployments—a critical capability gap for enterprises deploying autonomous systems.
The key facts
8 to knowNvidia Open Agent Safety Platform announced Monday Sep 28 2026
Claims to quarantine rogue agents within milliseconds
Uses OpenShell open-source software running on Vera AI CPU
Includes Sentry technology for monitoring
Platform allows users to restrict agent information access and validates restrictions before and during task execution
Response to wave of rogue agent hacking incidents reported by Reuters
No pricing, availability date, or operational limits disclosed
No independent testing or reliability data provided
Go to the source
The Verge AItheverge.com
Publisher excerpt: Nvidia is launching a new safety platform designed to contain and monitor AI agents, a move that comes in response to a wave of rogue hacking incidents, as reported earlier by Reuters. In an announcement on Monday, Nvidia says its new Open Agent Safety Platform can quarantine agents that attempt to…