From Pymnts · PYMNTS · · 1m 20s
Nvidia Builds AI Safety Watchdog for Banks
Nvidia launches hardware watchdog to stop AI agents from escaping controls.
In brief
Nvidia launches hardware watchdog to stop AI agents from escaping controls. Nvidia's Open Agent Safety Platform gives banks a hardware-based system to halt AI agents within milliseconds if they breach permissions. Originally reported by Pymnts.
Nvidia launches AI safety platform
Nvidia introduced Open Agent Safety Platform to let businesses stop autonomous ai agents even when software fails to keep them within bounds.
Hardware watchdog monitors ai behavior
Sentry, a separate watchdog running on Nvidia's BlueField-4 hardware, monitors ai agent behavior independently and can quarantine or stop it within milliseconds.
OpenShell enforces permissions
OpenShell creates a controlled environment for ai agents, traces their actions, and enforces operating rules to prevent unauthorized access.
Banks use the platform for safety
Over 100 organizations including banks like Citi and JPMorganChase are collaborating with Nvidia on shared open-source agent safety technologies.
SAP demonstrates real-world use
SAP's Joule Studio checks business authorization and employee roles to ensure ai agents only perform permitted actions within payment limits.
Critical security gap in ai agents
An ai agent could remain within technical permissions but propose inappropriate payments, or a valid payment could become dangerous if it had broad access.
Zero-trust approach for ai safety
The platform enables financial institutions to delegate narrow work while retaining the ability to inspect and interrupt ai agent execution.
The one thing
Nvidia's Open Agent Safety Platform gives banks a hardware-based system to halt AI agents within milliseconds if they breach permissions.
