Introducing AI Security

Use this topic to understand what AI Security is, what problems it solves, and what you can do with it before you begin setup.

Why AI Security matters

As AI tools become part of everyday work, they introduce new risks: sensitive data can leak through prompts and file uploads, security teams lose visibility into which AI tools are even in use, and autonomous agents can take actions on your organization's behalf without oversight. AI Security addresses this by helping you prevent sensitive data leaks, eliminate AI blind spots, and govern autonomous actions — all from a single platform.

What AI Security does for you

AI Security helps you take control of how AI tools and agents are used across your organisation. As more employees use tools like ChatGPT, Microsoft Copilot, and Claude — and as AI agents start accessing business data through applications like Salesforce and Atlassian — sensitive data can move in ways that are difficult to track or control. AI Security gives you the visibility, policy enforcement, and intelligent guidance to manage this risk.

AI Security is built on the Forcepoint Data Security Cloud platform — Data Loss Prevention (DLP), Data Security Posture Management (DSPM), Cloud Security Gateway (CSG), and Data Security Everywhere (DSE) — extending your existing data security policies to AI channels.

With AI Security you can:

  • Connect and monitor sanctioned AI apps — see who is using which AI tools, how often, and what data is involved. The dashboard shows users, agents, interactions, and DLP violations for each connected app.
  • Uncover Shadow AI — see which unapproved AI tools your employees are already using and who is using them, without deploying anything extra. Detection is derived from your existing web security traffic.
  • Protect sensitive data — apply DLP policies that detect and act on sensitive content moving through AI tools. The Sensitive Data Exposure tab shows violations by classification, by AI tool, and highlights the most critical events.

  • Measure policy effectiveness — the Policy Effectiveness tab shows how your DLP and access control policies are performing, which policies are generating the most hits, and where users are being impacted.

  • Govern AI agent traffic — use the MCP Gateway to inspect and control what your AI agents can read from or write to business applications. The AI Tool Landscape tab shows MCP Gateway Connections alongside your sanctioned apps.
  • Manage access control policies — create policies that allow, alert on, or deny specific agent or user activity.

  • Manage agents — register AI agents, organise them into groups, and track their activity. The Agent Management screen shows registered agents, active tokens, revoked tokens, and calls made today.

  • Get intelligent recommendations — ARIA, the Adaptive Risk Intelligence Assistant, guides you through setup, suggests policies, and surfaces risks as your environment evolves.

How AI Security is organised

AI Security is built around three pillars:

  • Sanctioned AI Governance — monitor and control usage of approved AI applications like Microsoft Copilot, ChatGPT, and Claude.
  • Agentic AI Governance — govern the tools and data that autonomous AI agents can access through the MCP Gateway.
  • Shadow AI Governance — discover and control use of unsanctioned AI applications and websites across your organization.
Note: Shadow AI Governance is not available in this Early Access (EA) release. It will provide visibility into unsanctioned AI apps and websites, with risk scoring and corporate-vs-personal account detection, in a future release.

Navigating AI Security

The Forcepoint Data Security Cloud Platform organises all AI Security features in the left navigation panel. Here is an overview of the two sections you will use throughout this collection.

AI Security

The AI Security section contains three menus:

1. Dashboard

Note: Cloud DLP integration is not available in this Early Access (EA) release. Data shown in the Sensitive Data Exposure and Policy Effectiveness tabs is placeholder data until this integration is enabled.

The AI Governance Dashboard is your central monitoring hub. For more information, see Read the AI Governance Dashboard. It has three tabs:

  • AI Tool Landscape — shows an overview of all sanctioned AI apps and MCP Gateway connections. For each app you can see the number of active users, agents, interactions, and DLP violations. Below this, the Unsanctioned AI Apps section lists the unapproved AI tools detected across your organisation, together with a Shadow AI Users count. Click any unsanctioned app to see which users accessed it. This data is derived from your web security traffic and needs no connector. See Understand How Shadow AI Detection Works.

  • Sensitive Data Exposure — shows total DLP violations, critical and high severity counts, users involved, and the top exposed data type. Violations are broken down by classification (Source Code, PII, Financial Data, Credentials, Customer Data) and by AI tool.

  • Policy Effectiveness — shows how your active policies are performing. You can see total hits, users impacted, escalations, and a breakdown of which DLP policies and access control policies are triggering most frequently.

2. Policy

Where you create and manage access control policies for MCP gateway connections. Policies are evaluated in priority order — the first matching policy is applied. From this page you can add, edit, delete, and reprioritise policies.

3. Agent Management

Where you review the AI agents active in your environment and the users behind that activity. See Dashboard chapter.

Where you register AI agents and organise them into groups. It has two tabs:

  • Agents — lists all registered agents with their owner, scope, token status, and last active date. From here you can register new agents and view their Client ID and Client Secret.

  • Agent Groups — lists all agent groups with the number of agents and description for each group. From here you can create new groups and assign agents to them.

Connector

The Connector section contains two menus:

1. App Connectors

Where you connect and manage sanctioned AI apps like ChatGPT, Microsoft Copilot, and Claude. From this page you can add a new app connector, view the live connection status of each app, and update credentials or settings for apps that are already connected.

2. Gateway Connectors

Where you connect business applications — such as Atlassian, Salesforce, and Microsoft 365 — to the MCP Gateway. From this page you can add a new gateway connector, configure its authentication method, select which tools are exposed to agents, and manage or disconnect existing connections.

Your AI Security journey

Journey 1 — Connect and monitor AI apps

Note: ARIA is not available in this Early Access (EA) release and may not appear in your tenant yet. Use this documentation for guidance in the meantime. ARIA will be enabled in a future release.

Start by going to the AI Security dashboard. A Configure AI Security dialog appears with two options:

  • Easy Setup — a guided three-step wizard that walks you through connecting your AI apps, setting up data protection policies with ARIA, and confirming your setup. If you skip data protection during setup, you can come back later via AI Security > Policy.
  • Configure Manually — go directly to Connector > App Connectors and configure each component at your own pace. Policies are managed separately via AI Security > Policy.

Both paths lead to the AI Governance Dashboard where you can monitor AI Tool Landscape, Sensitive Data Exposure, and Policy Effectiveness.

Journey 2 — Govern AI agent traffic (MCP Gateway)

After completing your AI apps setup, the What's Next cards on the finish screen guide you to configure the MCP Gateway. You can also access this anytime via Connector > Gateway Connectors.

  1. Add a gateway connector — Atlassian, Microsoft 365, or Salesforce
  2. Go to AI Security > Agent Management to register agents and create agent groups
  3. Go to AI Security > Policy to create and prioritise access control policies
  4. Monitor agent activity, tool calls, and DLP violations via the MCP Gateway dashboard