Key Terms
Definitions of the terms used throughout this guide.
Use this topic as a reference while you work through the rest of the guide.
Table 1. Terms used in this guide
| Term | Definition |
|---|---|
| Access policy | A rule that says whether an agent is allowed to perform a particular action in a business application. Policies can allow or deny. They are evaluated in priority order, and the first policy that matches is applied. |
| Adaptive Risk Intelligence Assistant (ARIA) | An assistant built into the platform. It appears alongside the pages you are working on and answers questions in context. ARIA is available from your first sign-in and needs no configuration. |
| Agent | An AI program that acts on your behalf, calling your business applications to read or change data without a person driving each step. Agents must be registered in AI Security before they can reach anything through the MCP Gateway. |
| Agent group | A named set of agents. Grouping agents lets you apply the same policy to several agents at once instead of writing one policy per agent. |
| Agentic AI Governance | The governance of AI agents acting on your business applications. Delivered by the AI Agent Gateway. |
| AI Agent Gateway | The capability that governs what AI agents can do in your business applications. In the platform, its features appear as the MCP Gateway. |
| App connector | The connection between AI Security and a sanctioned AI application such as ChatGPT Enterprise, Microsoft Copilot, or Claude Enterprise. It lets AI Security inspect what your users send to that application. |
| Client ID and client secret | The credentials issued to an agent when you register it. The agent presents these to prove its identity. The client secret is shown once, at the time you create it. |
| Data Loss Prevention (DLP) | Technology that detects sensitive data in content and applies an action such as allow, notify or block. In AI Security, DLP is delivered through data security policies that inspect what your people and agents share with AI tools. |
| Forcepoint Cloud Web Proxy | The Forcepoint product that monitors and classifies your organization's web traffic. Shadow AI reporting in AI Security depends on it. |
| Gateway connector | The connection between the MCP Gateway and one of your business applications, such as Atlassian. It lets AI Security control what agents can do inside that application. |
| MCP Gateway | The control point that all AI agent traffic passes through on its way to your business applications. Agents connect to the gateway rather than connecting to your applications directly. |
| Model Context Protocol (MCP) | An open standard that lets AI agents connect to business applications and data in a consistent way. |
| Monitor only mode | The state AI Security is in immediately after setup. Activity is recorded but no action is enforced. This lets you see what is happening before you decide what to block. |
| Sanctioned AI application | An AI application your organization has approved and connected to AI Security. |
| Sanctioned AI Governance | The governance of the AI applications your organization has approved. Delivered by AI Detection & Response. |
| Shadow AI | Use of AI applications your organization has not approved. AI Security identifies these from your existing web traffic rather than from a connector. |
| Tool | A single named action that a business application makes available to agents, such as creating an issue or reading a record. Access policies name individual tools, which is how you allow an agent to read from an application without letting it write to it. |
| Tool call | A single request from an agent to perform one tool action. |
| Unsanctioned AI Governance | The governance of AI applications your organization has not approved. Delivered by Shadow AI Discovery & Control. |