General-purpose AI chatbots are starting to look a bit passé. Your employees have moved on to the next chapter of AI: building and deploying AI agents. The good news? Most AI agents are built with existing infrastructure, in enterprise SaaS environments like Salesforce Agentforce, Microsoft Copilot Studio, and Notion AI, where modern work already happens.

‍

The bad news? AI agents often require highly permissive access to corporate data and tools to perform and automate tasks on employees’ behalf. At the current speed and scale of AI agent deployments, that’s already creating an unmanageable tangle of permissions and integrations that introduce unmitigated data exposure risks.

‍

In fact, 80% of organizations say they’ve encountered agentic AI risks, including improper data exposure and unauthorized system access, according to a recent Sailpoint survey.

‍

At this point, it’s safe to assume your workforce is building and deploying AI agents at work. What’s not safe to assume is that IT security and AI governance teams already have the visibility they need to answer questions like:

• Where are AI agents running?
• Who created each agent, and why?
• Who’s responsible for each agent if not the creator?
• What can AI agents connect to?
• What are AI agents able to do autonomously?
• What risks do these AI agents introduce?
• How do these AI agents impact our overall security posture?

Nudge Security answers these questions and more with our new AI agent discovery capabilities, released today in a research preview. Read on to learn how they work and how to get started.

‍

Introducing AI agent discovery from Nudge Security

Nudge Security’s AI agent discovery gives security teams visibility into the AI agents employees are building and deploying across enterprise SaaS platforms, along with their associated risks. It automatically discovers and inventories AI agents in your environment, connects each agent to its human creator, maps the integrations and data access permissions that power it, and surfaces risk insights based on its configuration and capabilities.

‍

Beyond discovery, you can set approval statuses, assign technical owners, and engage the right people through targeted nudges to understand why agents exist and ensure they’re used safely. It’s the first agentic AI governance solution built on the same deep SaaS and identity context that already powers Nudge Security’s AI and SaaS security platform.

‍

Unlike point solutions built solely to secure AI agents, Nudge Security brings deep visibility of access, integrations, identity signals, and behavioral context that reveals how employees are actually using new SaaS and AI technology across the enterprise. Our new AI agent discovery capabilities extend this foundation directly into the agentic layer, with no new deployments required for existing customers already connected to SaaS environments like Salesforce and ServiceNow.

‍

Key features

• Discover AI agents across agentic platforms. Get continuous visibility into agents running across Microsoft Copilot Studio, Salesforce Agentforce, ServiceNow, n8n, Tines, Workato, Abacus, ChatGPT, and Gemini for Google Workspace.
• Inventory agent permissions, resources, and capabilities. For every agent in your environment, you’ll know what it can access, what it’s authorized to do, and who built it.
• Surface agentic risks. Risk insights and security posture findings highlight the issues that matter: publicly accessible agents, hardcoded credentials, unauthenticated MCP connections, high-risk integrations, and agents that have outlived their creators.
• Enforce AI agent guardrails. Reach the people responsible for agents directly, and give them a clear path to confirm intent, justify access, and resolve issues before they compound.

‍

‍

How to get started

Nudge Security’s AI agent discovery takes advantage of our connected apps infrastructure, available to customers and trial users. Connected apps that support AI agent discovery are marked with a badge in the product. For customers already connected to environments like Salesforce and ServiceNow, there are no additional configurations or deployments needed. Simply request access to join our research preview. New connections for additional agentic platforms are available for ChatGPT, Microsoft Copilot, and Gemini.

‍

Nudge Security is introducing AI agent discovery as a research preview for testing, feedback, and real-world experimentation, similar to how other leading technology providers like Anthropic provide access to early versions of new capabilities.

‍

AI security and governance with Nudge Security

AI agent discovery is the latest addition to Nudge Security’s growing set of AI security governance capabilities.

‍

Earlier this year, we launched detection for remote MCP server connections, giving security teams visibility into the OAuth grants that power AI agents. We also launched AI data flow visualization and sensitive data sharing detection, alongside Day One discovery of shadow AI apps, users, and integrations across thousands of AI providers.