ServiceNow and Nvidia Launch Arc AI Agent With Layered Security Checks

What happened

ServiceNow and Nvidia have launched Arc, a new AI agent for enterprise customers, according to an exclusive report from ZDNET. The partners are positioning the product around a specific problem: how to keep AI agents secure once they are working inside business systems.

Their stated approach is to use multiple layers of real-time security checks rather than a single control. ZDNET describes this as the companies' attempt to resolve what it calls the AI agent security conundrum.

The available reporting covers the launch and its security positioning. Pricing, availability and technical architecture are not detailed in the material we reviewed, so readers should check the original report before relying on those points.

Why it matters

Enterprise AI is moving from systems that answer questions to systems that take action. An agent can read records, trigger workflows and complete tasks on a customer's or employee's behalf. That shift changes the stakes: a security weakness in an agent can become a service failure, such as an incorrect approval, an exposed record or an unintended change to a customer's account. Launches that foreground security in this way suggest that vendors see trust as a prerequisite for enterprise adoption, not a feature to add later.

For leaders in digital transformation, the practical question is less whether agents can be deployed and more which decisions they should be allowed to make, under what conditions, and with what oversight. Layered, real-time checks imply that governance has to operate while the work is happening, not only during design reviews or after an incident. Organisations planning agent programmes may need to map which actions require human confirmation, how those controls appear to users, and how they will be monitored over time.

The Renascence take

Most coverage of AI agents will focus on capability, but the experience question is about permission. Every checkpoint an agent passes is a moment where a customer or employee either trusts the system or is quietly exposed to it.

Security controls are often treated as invisible plumbing, yet for an AI agent they shape the experience itself: what it refuses to do, when it pauses, and how clearly it explains either. The behavioural principle underneath is that people accept automation when its boundaries are legible and predictable. A customer-obsessed operator should therefore map agent actions by risk, design visible confirmation moments for consequential decisions, and measure the trust outcomes of those moments rather than only task completion.