Nvidia open platform shifts AI agent security to external layer

Nvidia released an open platform that externalizes security controls for AI agents, moving the lock from inside the model to outside. This is a technical shift in AI security architecture, but the author notes that external locks do not equate to trustworthiness. The development signals a trend toward standardized external guardrails for AI agents.