ServiceNow and Nvidia Just Solved the AI Agent Security Nightmare

According to ZDNET reporting, companies have largely hesitated to adopt fully autonomous AI systems following a wave of agent-led security breaches over the past year. To counter this, ServiceNow and Nvidia have teamed up to release Arc, a persistent enterprise AI agent built specifically to run locally on Nvidia's new RTX Spark hardware while maintaining strict administrative guardrails.

How security works in silicon

Unlike standard software-level chatbots, Arc shifts enforcement directly into the runtime and hardware layers. The system relies on the Super Intelligence Control Tower from ServiceNow to monitor agent activity, vet access privileges, and check for vulnerabilities like prompt injection.

When an infraction occurs, the control tower revokes the agent's digital AI Passport. According to the announcement, Nvidia BlueField hardware can enforce this revocation at machine speed before the agent can execute unauthorized actions, containing risky tasks inside a secure sandbox without crashing the entire application workflow.

How to run and manage Arc

Arc is generally available now for enterprise deployment, though platform availability varies by operating system:

• Hardware setup: Designed to run efficiently on Nvidia RTX Spark devices, keeping sensitive enterprise data local whenever possible.
• OS support: Currently available on Mac computers, with official Windows support scheduled for release soon.
• Policy enforcement: IT administrators configure security rules and risk tolerances directly inside the ServiceNow Super Intelligence Control Tower.

Is Arc right for your workflow?

If your team handles sensitive internal communications and relies on productivity tools like email and Microsoft Teams, managing autonomous software carries inherent risks. Arc's ability to offload complex multi-step tasks to frontier models only when necessary reduces latency, while hardware-level sandboxing offers peace of mind.

However, early adopters must keep hardware prerequisites in mind. Because Arc depends heavily on specialized Nvidia devices and is currently optimized primarily for Mac environments, Windows-heavy offices will need to wait for full deployment support.