Review discovered AI tools

AI Discovery detects and classifies the AI tools actually used in your organization. Your job as an admin or security lead is triage: decide, tool by tool, what is sanctioned, what needs review, and what gets blocked.

Newly discovered vs. catalog

Discovery has two tabs: Newly discovered — tools recently observed by Shield Web that are not yet approved in your catalog — and Catalog, your reviewed inventory. When unreviewed tools appear, Shield prompts you: "New AI tools detected — please triage them."

Every app row shows vendor, functional category (e.g. LLM chat, coding assistant), when it was first seen, how many users it affects, and its source (Verified by the Qadar catalog, auto-detected, or manually added).

Risk and status

Two dimensions drive the triage:

• Risk — based on data exposure, vendor trust, and compliance posture. The KPI tiles let you jump straight to high-and-critical-risk or ungoverned apps.
• Status — the approval state in your organization: sanctioned, under review, unsanctioned, or blocked.

Sanction, review, or block

Use the quick actions on each app — also in batch:

• Allow — marks the app as sanctioned for your organization.
• Warn / Review — flags it for evaluation; the app stays accessible while your team decides.
• Block — employees are prevented from using it.

A blocked or allowed status takes effect through Shield Web like any policy. For finer rules on a sanctioned tool — specific data types, specific actions — create a policy scoped to that tool.

Internal tools

Self-hosted or internal AI (an Azure OpenAI deployment, a custom RAG app) won't be auto-detected — use Add custom app so it appears in your catalog and governance picture like everything else.

Tip: the "Ungoverned" metric (apps in review or unsanctioned) is the number to drive to zero — it is your live shadow-AI backlog.