Data Security for AI | Overview

The Data Security for AI solution enables organizations to monitor and control the use of both approved AI and shadow applications. It delivers AI-specific detection, governance, and remediation capabilities to help reduce the risk of sensitive data flowing into approved and public LLMs.

As generative AI tools are rapidly adopted across teams—often without centralized oversight—organizations face increased risk, including sensitive data leakage through prompts and AI-generated responses, and the use of unsanctioned (“shadow”) AI applications that bypass existing controls.

Data Security for AI mitigates these risks by enforcing data protection policies and providing visibility into how AI tools are accessed and used. It also strengthens the overall security posture of AI tools through SaaS Security Posture Management (SIPM), which is included with this solution.

Data Security for AI requires a separate license. Users will automatically receive SIPM product.

Currently Data Security for AI is supported for Copilot (MS), ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google).

Use Cases

Following are use cases for Data Security for AI:

Prevent sensitive data exposure: Enforce data protection policies by automatically detecting and redacting sensitive data in prompts before they are submitted to unsanctioned AI applications.

Monitor sanctioned AI usage: Detect when prompts submitted to sanctioned AI tools contain potentially harmful, risky, or policy-violating content, enabling investigation and remediation of insider misuse or abuse.

Protect against data leakage in AI responses: Assess risk and take corrective action when responses from sanctioned AI applications contain sensitive or regulated data.

Identify and reduce unapproved AI risk: Discover unsanctioned AI applications in use within the organization, including those with third-party access to your organization's data.

Data Security for AI

The Data Security for AI app includes:

• Data Security for AI | Dashboard: Provides a 3-step process to manage AI risk across your organization. It highlights key exposure areas and helps you quickly identify where sensitive data may be overexposed.

• Data Security for AI | Sensitive Data in Unapproved AI: Highlights AI applications that are not sanctioned or formally approved, along with the potential risk of unmanaged data exposure.

• Data Security for AI | Sensitive Data in Approved AI: Highlights AI applications that are sanctioned or formally approved and their exposure.

• Data Security for AI | Approved AI Posture: Provides a view of approved AI applications, guiding you through the process of security posture improvement.

• Data Security for AI | Activities: Helps you understand where the data comes from.

Data Security for AI Notes

For Co-Pilot only: When attaching a file to prompt that, the message File Upload to Generative AI does not display for File Access events. Files from OneDrive, SharePoint, or previously used files are classified as File Access events. Only local (real) files qualify for a File Upload to Generative AI event.