NVIDIA and ServiceNow Launch Project Arc Desktop AI Agent

NVIDIA and ServiceNow unveiled Project Arc, an enterprise autonomous desktop agent secured by NVIDIA OpenShell and governed by ServiceNow AI Control Tower.

NVIDIA and ServiceNow Launch Project Arc Desktop AI Agent

NVIDIA and ServiceNow have expanded their partnership to launch Project Arc, an enterprise autonomous desktop agent designed to live on employee machines and complete complex, multi-step work on its own — under governance strict enough for cautious enterprise security teams to approve.

An agent that acts, not just answers

Unveiled at ServiceNow's Knowledge 2026 conference, Project Arc is a long-running, self-evolving agent that thinks, writes code, executes and adapts when tasks go off-script. It can reach into local file systems, terminals and installed applications to handle work that traditional automation cannot, completing jobs across enterprise tools without pre-built workflows.

NVIDIA accelerated computing infrastructure powering enterprise AI agents

Governed and secure from day one

Every action Project Arc takes runs inside NVIDIA OpenShell, an open-source sandboxed runtime that contains autonomous activity under policy-based controls. ServiceNow AI Control Tower governs those actions — setting policies, monitoring behavior and logging every file read, command executed and API called. Grounded in ServiceNow's Action Fabric and configuration database, the agent draws on deep context about how an organization actually works.

Benchmarks and AI factories

The companies are also advancing NOWAI-Bench, an open benchmarking suite for enterprise agents, and extending governance into the data center through the NVIDIA Enterprise AI Factory validated design. NVIDIA says its Blackwell platform delivers more than 50x greater token output per watt than Hopper, sharply lowering the cost of running always-on agents at scale.

The move deepens a partnership that has tracked the broader enterprise-AI build-out, from NVIDIA's acquisition of Kumo AI to physical-AI tie-ups like FANUC and Google's Gemini collaboration and the race for custom inference silicon seen in OpenAI and Broadcom's first inference chip. Project Arc is available as an early preview.

Reporting based on coverage from the NVIDIA Blog and ServiceNow Newsroom.

Category: AI & Technology

Tags: AI integration AI Automation Enterprise AI artificial intelligence AI Foundation Models

Related Articles