ServiceNow and Nvidia have released Arc, a long-running personal AI agent built for everyday enterprise work, the two companies said Wednesday. The agent is generally available now on Mac, with Windows support still in development. The companies first announced their plans for the agent in May.
Arc is aimed at daily tasks handled through productivity tools such as email and Microsoft Teams. It is designed to run efficiently on Nvidia's new RTX Spark devices and can operate locally on an employee's desktop or on its own device within Nvidia OpenShell, an open-source tool that controls runtime. ServiceNow and Nvidia said Arc accumulates memory and skills over time so its performance continues to improve. It also scans an employee's inbox and meeting notes to surface work and action items proactively, and it hands more complex tasks to frontier models only when necessary, keeping data as local as possible otherwise.
The companies positioned security as what distinguishes Arc from other enterprise agents. ServiceNow's Super Intelligence Control Tower governs and secures the agent while Nvidia OpenShell enforces controls at runtime, allowing Arc to work independently while the enterprise retains control. The Control Tower monitors agent activity, vets access, checks for risks such as prompt injection and measures agent performance. ServiceNow and Nvidia said enforcement extends beyond the software layer into the runtime and hardware layers, which they described as a more reliable approach.
IT teams define policies in the Super Intelligence Control Tower, and Nvidia OpenShell can be directed to enforce them at runtime. ServiceNow and Nvidia said users get granular control over which systems agents can access, which APIs they can call and which data they can read.
The security design also includes AI Passport, which the companies announced the prior week. It holds an enterprise's rules, set by the Control Tower, on risk tolerance and the services Arc and other agents may access, and it acts as a bridge between ServiceNow infrastructure and Nvidia's checkpoints so OpenShell can intervene when an agent attempts a disallowed action.
If the Control Tower revokes an agent's passport, OpenShell confines the attempted task to a sandbox for later review by auditors while other approved actions continue without a restart. Nvidia BlueField can then enforce the revocation at machine speed before the agent proceeds, which the companies said creates enforcement in silicon intended to prevent overrides from compromised software. The system is designed to continuously monitor agent actions and permissions to prevent breaches and unsanctioned actions.
More AI news from TechManNews.
