ServiceNow and NVIDIA launch Arc, a personal AI agent for the enterprise with layered security

ServiceNow and NVIDIA have made Arc generally available, an AI agent both companies describe as "a long-running personal AI agent that can take on everyday work." According to ZDNET, the idea was originally unveiled in May and the product became available on a Wednesday in October 2026. For now it runs on Mac; Windows support is still on the way.

The context the article lays out is one of the main obstacles to enterprises adopting more autonomous AI systems: adequate security controls. ZDNET notes that this has worsened after a summer of agent-driven breaches. ServiceNow has partnered with NVIDIA to address that need by combining AI capabilities with safeguards designed with security in mind from the outset.

What Arc does: like other enterprise agents, it is designed for daily tasks that rely on productivity tools such as email and Microsoft Teams. It is optimized to run efficiently on the new NVIDIA RTX Spark devices. It can run locally on an employee's desktop or inside NVIDIA OpenShell on their own device. The companies say Arc builds up memory and skills over time, so its performance continuously improves. They also say it is efficient: it hands off the most complex tasks to frontier models only when necessary and keeps data as local as possible otherwise. In addition, according to both companies, it "proactively surfaces pending work and actions" by scanning the employee's inbox and meeting notes.

The real differentiator, according to the companies, is not so much the capabilities as the multi-layered security framework surrounding Arc. Both claim their approach delivers more reliable security by "moving enforcement beyond the software layer, into the runtime and hardware layers."

Governance falls to the ServiceNow Super Intelligence Control Tower, ServiceNow's AI agent governance framework, conceived as an end-to-end service that monitors what agents do, checks their access, detects security risks such as prompt injection and measures agent performance, among other functions. NVIDIA OpenShell, an open-source tool that controls the runtime environment, is responsible for enforcing those policies. In the words of the announcement, IT defines policies in the Control Tower and OpenShell can enforce them at runtime; users gain granular control over which systems agents can access, which APIs they can call and which data they can read.

Another layer is AI Passport, announced by both companies the previous week. It contains the company's set of rules, established from the Control Tower, on risk tolerance and the services that Arc and other agents can access. It acts as a bridge between ServiceNow's infrastructure and NVIDIA's enforcement points, and allows OpenShell to step in efficiently if an agent takes an action it shouldn't.

If the Control Tower tells OpenShell that an agent's passport has been revoked, OpenShell will keep the task the agent was trying to execute contained in a sandbox, where reviewers can audit it later. The interruption is task-specific, so the agent's other approved actions can continue without restarting. In addition, according to the companies, when a passport is revoked, NVIDIA BlueField can enforce that revocation "at machine speed" before the agent proceeds, creating "in-silicon" enforcement designed to prevent compromised software from overriding it.

Taken together, the system is designed to continuously monitor the agent's actions and permissions in order to prevent breaches, unauthorized actions and other consequences of using agents. It is worth bearing in mind that the claims about this architecture's reliability come from the companies themselves; the article provides no independent performance data or figures on pricing or customers.