---
name: Copilot Notebooks Data Governance Review
description: Reviews shared Copilot Notebooks across the tenant — grounding sources connected, sensitivity label inheritance on notebook outputs, guest and external access controls, and data residency compliance for Australian-regulated content.
---

# Copilot Notebooks Data Governance Review

> **TL;DR:** This skill reviews shared Copilot Notebooks for grounding source sensitivity, whether notebook outputs inherit sensitivity labels, whether OFFICIAL:Sensitive content is accessible to external guests, and whether data residency requirements for Australian-regulated content are met.

## What does the Copilot Notebooks data governance review cover?

Copilot Notebooks are shared project workspaces in Microsoft 365 that ground the Copilot AI on a curated collection of content: SharePoint documents, OneDrive files, Teams meeting recordings, OneNote pages, and external URLs. Notebooks synthesise that content into outputs — summaries, Excel spreadsheets, infographics, action plans — within a shared workspace that team members (and potentially external guests) can access. This creates a data governance challenge: if a notebook is grounded on Highly Confidential SharePoint documents, does a sensitivity label carry through to the notebook's AI-generated outputs? Can an external guest invited to the notebook access summaries derived from OFFICIAL:Sensitive source material? Do notebooks respect Australian data residency requirements for regulated content? This skill reviews all shared Copilot Notebooks across the tenant and assesses these governance dimensions.

## When should you run this skill?

- "Audit shared Copilot Notebooks for sensitive content exposure"
- "Check whether notebook outputs carry sensitivity labels from grounding sources"
- "Identify notebooks with external guest access connected to sensitive content"
- "Review Copilot Notebooks for PSPF or OFFICIAL:Sensitive compliance"
- "Verify data residency for notebooks used in Australian government or regulated contexts"
- "Produce notebook governance evidence for IRAP assessment"

## How this skill works, step by step

1. Identify the Microsoft 365 groups and shared workspaces that contain Copilot Notebooks: use Graph API `GET /groups/{id}/notebooks` or search the M365 Copilot admin centre for notebook activity
2. For each shared notebook: record the notebook name, owning group or workspace, creation date, last modified date, and member list
3. Identify external or guest members: query the notebook's parent group for guest accounts (`userType eq 'Guest'`)
4. Enumerate grounding sources connected to each notebook: navigate to the notebook configuration and list all connected knowledge sources (SharePoint libraries, OneDrive folders, Teams meeting recordings, external URLs)
5. For each grounding source, determine the highest sensitivity label present in that source:
   - SharePoint library: check the library's default label and the highest label on any document in scope
   - OneDrive folder: check file-level sensitivity labels
   - Teams meeting recording: check the label applied to the recording file in OneDrive
6. Assess sensitivity label inheritance on notebook outputs: when Copilot generates an Excel spreadsheet, Word document, or summary from notebook content, does the output inherit the highest sensitivity label from the grounding sources? Check the M365 Copilot label inheritance policy configuration
7. Flag notebooks where the notebook has external guest members AND the grounding sources include Confidential, Highly Confidential, or OFFICIAL:Sensitive content
8. Check data residency: for tenants with Multi-Geo or Australian data residency requirements, verify that notebook data (grounding content, AI-generated outputs) is stored in the Australian geography
9. Review notebook sharing settings: can notebook members share the notebook or its outputs externally? Check whether the parent group's external sharing policy allows this
10. Produce the notebook inventory and governance summary below

## Output format

### Notebook inventory

| Notebook name | Owner group | Grounding sources | Highest source label | External guests | Label inheritance enforced | Data residency | Risk |
| --- | --- | --- | --- | --- | --- | --- | --- |
| Q3 Strategy Review | Executive Leadership | ExecComms SP library, Teams recordings | Highly Confidential | 2 guests | Not confirmed | AU East | Critical — HC content + external guests |
| Project Phoenix | IT Engineering | IT-Projects SP library | General | None | N/A | AU East | Low |
| HR Onboarding Guide | People & Culture | HR-Policies library | OFFICIAL:Sensitive | None | Yes | AU East | Medium — verify label inheritance on outputs |

### Governance summary

- Total shared Copilot Notebooks in scope: N
- Notebooks with external guest members: N
- Notebooks with Confidential or above grounding sources: N
- Notebooks with external guests AND sensitive grounding sources: N (Critical)
- Notebooks where sensitivity label inheritance on outputs is not confirmed: N
- Notebooks with data residency outside Australian geography: N
- Recommended actions: remove external guests from notebooks grounded on Confidential or above sources; confirm label inheritance policy covers notebook-generated Excel, Word, and image outputs; review multi-geo data placement for regulated content notebooks

## Sensitivity label inheritance — current Microsoft 365 behaviour (June 2026)

Copilot Notebooks inherit the highest sensitivity label from grounding sources for some output types (Word documents, Excel spreadsheets generated via the Excel agent) when the tenant label inheritance policy is enabled. Image outputs (infographics) and inline notebook summaries may not inherit labels in all configurations. Verify your tenant's label inheritance policy in the Microsoft Purview portal → Information Protection → Label policies → Auto-labelling settings.

## Scope and safety

Read-only — this skill does NOT:

- Modify notebook grounding sources or content
- Change sensitivity label assignments on notebooks or outputs
- Remove external guest access
- Alter data residency or Multi-Geo settings

## Licensing and permissions

### Licences and add-ons

| Capability used | Minimum licence |
| --- | --- |
| Copilot Notebooks (shared notebooks) | Microsoft 365 Copilot licence per user |
| Sensitivity label inheritance on Copilot outputs | Microsoft 365 E5 Compliance or Microsoft Purview Information Protection add-on |
| Multi-Geo data residency verification | Microsoft 365 Multi-Geo add-on (for organisations requiring Australian-only data residency) |

### Least-privilege roles

- Global Reader or Reports Reader (read Microsoft 365 group membership and sharing settings)
- Compliance Administrator (read sensitivity label policies and label assignments)
- SharePoint Administrator (verify data residency and external sharing policies)

### Microsoft Graph permissions (read-only)

- `Group.Read.All` — enumerate Microsoft 365 groups and their members (including guests) that own Copilot Notebooks
- `Sites.Read.All` — read SharePoint grounding source library content and sensitivity labels
- `Files.Read.All` — read OneDrive and SharePoint file-level sensitivity label assignments
- `InformationProtectionPolicy.Read.All` — read label inheritance policy configuration and label definitions

## Sources and compliance

- [What's New in Microsoft 365 Copilot — May 2026 (Microsoft Tech Community)](https://techcommunity.microsoft.com/blog/microsoft365copilotblog/what%e2%80%99s-new-in-microsoft-365-copilot--may-2026/4522010)
- [Microsoft 365 Copilot release notes (Microsoft Learn)](https://learn.microsoft.com/en-us/microsoft-365/copilot/release-notes)
- [Sensitivity labels in Microsoft 365 Copilot (Microsoft Learn)](https://learn.microsoft.com/en-us/purview/sensitivity-labels-copilot)
- Aligns with PSPF Requirement 4 — OFFICIAL:Sensitive content accessed by AI tools must be controlled and not accessible to unauthorised personnel (including external guests)
- Supports the Privacy Act 1988 (Cth) — AI-generated summaries derived from personal information inherit the privacy obligations of the source data
- Aligns with IRAP evidence requirement for data handling and information security controls on AI-generated content
- Pair with Purview AI Activity Explorer to correlate notebook AI interactions with the activity explorer event stream
- Output in Australian English