Exclusive: ServiceNow and Nvidia launch Arc, a new AI agent for enterprise
Original reporting by ZDNet
Arc is a personal, long-running AI agent developed through a partnership between ServiceNow and Nvidia, designed to automate and streamline everyday work tasks for businesses. Now generally available, Arc integrates seamlessly with common enterprise productivity tools like email and Microsoft Teams, running efficiently on Nvidia’s new RTX Spark devices either locally or within Nvidia OpenShell. While Arc’s core mission is to empower employees by handling routine tasks and continuously improving its performance through accumulated context, its most significant innovation lies in its comprehensive, multi-level security architecture.
Prioritizing AI security
The widespread adoption of autonomous AI systems has been hampered by concerns over adequate security controls, especially in the wake of recent agent-led breaches. ServiceNow and Nvidia directly tackle this roadblock by engineering Arc with security as a foundational principle. Their framework extends enforcement beyond the typical software layer, integrating it into runtime and hardware. This robust system combines ServiceNow’s Super Intelligence Control Tower, which defines granular company policies, with Nvidia OpenShell, restricting agent behavior and enforcing these directives at runtime. An additional layer, AI Passport, further specifies an enterprise’s risk tolerance and permissible agent actions. This holistic, "enforcement in silicon" approach provides continuous monitoring and the ability to prevent unsanctioned actions, giving businesses greater confidence in deploying powerful AI agents.
Arc's general availability on Nvidia RTX Spark devices signals a significant evolution in enterprise AI agents. While its productivity-enhancing features for daily tasks are noteworthy, the true differentiator lies in its multi-layered security framework. By integrating ServiceNow's Super Intelligence Control Tower, Nvidia OpenShell, and the innovative AI Passport system, Arc directly addresses the persistent anxieties surrounding autonomous AI, particularly in the wake of recent agent-led breaches. This robust architecture, designed for continuous monitoring and hardware-enforced policy adherence, aims to instill enterprise confidence by shifting security enforcement beyond software into the runtime and silicon layers, thereby minimizing risk at a foundational level.
A new security paradigm
This collaboration between ServiceNow and Nvidia represents more than just a product launch; it’s a strategic response to one of the most significant roadblocks to enterprise AI adoption: trust and control. By offering granular oversight and a proactive defense mechanism against unauthorized actions, Arc sets a formidable new benchmark for agent security. Its success could accelerate the widespread integration of autonomous AI agents into critical business workflows, enabling organizations to leverage AI’s full potential without compromising data integrity or operational security. This move underscores a maturing AI landscape, where the convergence of powerful agents and sophisticated security protocols will redefine how enterprises safely automate and innovate. The precedent set by Arc’s comprehensive security model is likely to influence future AI agent development, prompting other vendors to prioritize and build in similar robust safeguards, ultimately paving the way for a more secure and pervasive AI-driven enterprise future.
Frequently asked questions
- What is Arc, the new AI agent, and what are its primary functions?
- Arc is a personal, long-running AI agent developed by ServiceNow and Nvidia, now generally available. It assists employees with daily productivity tasks like managing emails and Microsoft Teams. Arc is designed to learn and improve over time, proactively surfacing work and action items while efficiently running on Nvidia's new RTX Spark devices, keeping data local when possible.
- How does the Arc AI agent ensure robust security for enterprise data?
- Arc features a multi-level security framework that enforces policies at software, runtime, and hardware layers. ServiceNow's Super Intelligence Control Tower defines rules, while Nvidia OpenShell restricts agent behavior. An "AI Passport" holds enterprise risk tolerance rules, allowing Nvidia BlueField to enforce revocations rapidly in silicon, containing unauthorized actions and preventing breaches.
- What are the hardware requirements and availability for the Arc AI agent?
- Arc is designed to run efficiently on Nvidia’s new RTX Spark devices, either locally on an employee's desktop or within Nvidia OpenShell on its own device. The AI agent is currently generally available for Mac users. Support for Windows devices is planned and will be released at a later date, expanding its platform accessibility.