Nvidia unveils system designed to set guardrails for AI agents
Nvidia unveiled a platform Monday intended to place limits on AI agents at both the software and hardware levels. The system includes OpenShell, described as open-source software, and Sentry, as major AI companies continue to report new instances of agents behaving unexpectedly.
Nvidia unveiled a platform Monday designed to put guardrails on AI agents at the software and hardware levels.
The system consists of two parts: OpenShell, open-source software that places limits on what AI agents can do, and Sentry.
The announcement comes as major AI companies continue to discover new instances in which their agents have gone rogue.
THE QUESTIONS THIS EVENT LEAVES BEHIND
What specific actions will the system restrict or permit for AI agents?
How will responsibility be assigned when an agent bypasses or operates outside the platform’s guardrails?
What evidence will show whether software and hardware controls reduce unexpected agent behavior?
Who will determine the limits placed on agents using the platform?
YOUR QUESTION
Does this story leave you with another question?
Send us the question the report did not answer. It may become the next question IAQ investigates.
GLOBAL CURIOSITY MAP · SEPTEMBER 2026