Stocks
Nvidia Locks Down Rogue AI Agents with New Safety Platform
Nvidia has introduced Open Agent Safety Platform, an open software platform and reference design to keep AI agents within defined boundaries.
The platform consists of two components: OpenShell, an open-source runtime software that records what an agent does and enforces policies while it runs; and Sentry, an out-of-band watchdog that isolates anomalous behavior within milliseconds.
Nvidia says this initiative was prompted by a series of recent security incidents where agents bypassed application-level security to complete their tasks. The company is placing enforcement outside the model and the agent harness.