Control 4.1: Copilot Admin Settings and Feature Management

Control ID: 4.1 Pillar: Operations & Monitoring Regulatory Reference: Sarbanes-Oxley §§302/404, FFIEC IT Examination Handbook, GLBA §501(b) Last Verified: 2026-05-25 Governance Levels: Baseline / Recommended / Regulated

Scope boundary: FSI-CopilotGov vs FSI-AgentGov

This control governs the Microsoft 365 Copilot surface only — tenant-level configuration, data-source posture, audit/eDiscovery, and admin-managed extensibility. Governance of the agents themselves (Copilot Studio agents, declarative agents, Agent Builder, custom pro-code agents) — including agent registration, risk tiering, environment zoning, model-card review, and lifecycle promotion — lives in the companion FSI-AgentGov framework. See Relationship to FSI-AgentGov for the full boundary map.

Objective

Establish centralized governance over Microsoft 365 Copilot administrative settings, role assignments, feature availability, and adjacent billing or cloud-policy decisions. This control helps organizations manage Copilot consistently across Microsoft 365 without relying on ad hoc administrator judgment or overly privileged roles.

Why This Matters for FSI

Financial institutions operate under internal control frameworks that require documented, auditable, and role-separated management of technology configuration changes. Sarbanes-Oxley §§302/404 expects management to maintain and test IT general controls where technology affects regulated reporting and operational processes. The FFIEC IT Examination Handbook similarly emphasizes access governance, change control, and ongoing monitoring for enterprise technology platforms.

For Copilot, those expectations apply to more than one settings page. Administrators now work across the Copilot overview, Copilot settings tabs, Agents administration, billing controls, and Cloud Policy. If these surfaces are managed inconsistently, firms can end up with unauthorized feature rollouts, unmanaged agent exposure, or unexpected spend paths that weaken governance evidence and change control discipline.

Least-privilege administration also matters. Microsoft now supports AI Administrator as the primary low-privilege role for Copilot scenario management in the Microsoft 365 admin center, while Global Reader supports read-only review. Overuse of broad administrator roles creates unnecessary risk in regulated environments.

Disclaimer

This control is provided for informational purposes only and does not constitute legal, regulatory, or compliance advice. See full disclaimer.

Control Description

Microsoft 365 Copilot administration now spans a small set of connected governance surfaces:

Surface	Primary Path	Governance Use
Copilot overview	M365 Admin Center > Copilot > Overview	Readiness, usage trends, security links, and recommended actions
Copilot settings	M365 Admin Center > Copilot > Settings	Tenant controls grouped under User access, Data access, Copilot actions, and Other settings
Agents administration	M365 Admin Center > Agents > Overview / All agents / Settings	Agent inventory, lifecycle, user access, allowed types, and sharing
Agent 365	M365 Admin Center > Agents (centralized platform)	Unified agent registry, cross-platform telemetry, and lifecycle governance across M365, Copilot Studio, and third-party agents
Copilot security pivot	M365 Admin Center > Copilot > Security	Policy creation to limit Copilot data access scope, restrict grounding sources, and enforce data boundaries per user or group
Billing and PAYG	M365 Admin Center > Billing > Pay-as-you-go services / Cost Management	Billing policies, budgets, and metered usage monitoring
Cloud Policy	`https://config.office.com` > Customization > Policy Management	Copilot Pages / Notebooks and code preview policy decisions
Baseline Security Mode	M365 Admin Center > Settings > Org settings > Security & privacy	Foundational Microsoft 365 security baseline that complements Copilot-specific controls

Role Separation

Administrative Area	Recommended Role	Why
Copilot scenario and settings review	AI Administrator	Least-privilege role for Copilot and agent administration in the Microsoft 365 admin center
Read-only evidence collection	Global Reader	Supports oversight without configuration rights
Org-wide exception handling	M365 Global Admin	Reserve for emergency or broader tenant changes that require elevated privilege
Teams-specific Copilot controls	Teams Admin	Owns meeting and messaging policy surfaces
Outlook-specific Copilot controls	Exchange Online Admin	Owns mailbox and Exchange-related settings
SharePoint / SharePoint Embedded review	SharePoint Admin	Owns storage, sharing, and container administration
Cloud Policy for Copilot app experiences	Office Apps admin or approved equivalent	Manages Pages/Notebooks and other app-level cloud policies

What Should Be Governed Together

This control treats the following decisions as part of one operational governance process:

who can access Microsoft 365 Copilot
whether web search and other data-access features are enabled
whether agents are allowed, restricted, or limited to approved groups
whether third-party model providers are enabled or restricted for agents
whether PAYG billing is enabled for approved users or departments
whether Copilot Pages and Copilot Notebooks are available through Cloud Policy
whether Copilot data access scope has been restricted through the Copilot security pivot
whether change evidence exists for each configuration adjustment

Baseline Security Mode

Baseline Security Mode (BSM) is an opt-in, secure-by-default configuration that covers 18–20 settings across Office, Exchange, Teams, SharePoint, and Entra. BSM is relevant to Copilot governance as a complementary foundation, but it should not be documented as a Copilot-specific tab. Microsoft documents it as an organization-wide Microsoft 365 security baseline under Settings > Org settings > Security & privacy > Baseline Security Mode.

Portal: M365 Admin Center > Settings > Org settings > Security & privacy > Baseline Security Mode

BSM Area	Key Settings	Copilot Governance Relevance
Authentication	Block legacy authentication protocols; mandate phishing-resistant MFA	Reduces the risk of unauthorized Copilot access through credential compromise
Files	Restrict legacy file types; disable ActiveX controls	Limits Copilot grounding exposure to file formats with known security risks
Room Devices	Secure shared device configurations for meeting rooms	Supports governance of Copilot in shared-device meeting scenarios

BSM includes a simulation mode that enables administrators to analyze the impact of enforcement before activating settings. Organizations should run simulation mode first to identify any workflows or integrations that would be affected, then transition to enforcement after documenting the impact assessment.

BSM availability

Baseline Security Mode reached general availability in November 2025. For FSI teams, BSM is best treated as a complementary control surface that strengthens the broader Copilot security posture rather than replacing workload-specific Purview, SharePoint, or agent governance decisions.

Entra AI Administrator Homepage

Microsoft provides a dedicated AI admin homepage in the M365 Admin Center for users assigned the Entra AI Administrator role. The homepage consolidates AI-specific administration into a single starting point:

At-a-glance AI metrics — Copilot adoption, usage trends, and security posture indicators
Quick links — Direct navigation to Copilot settings, agent governance, DLP policies, and DSPM for AI
Curated resources — Links to Microsoft Learn documentation, best practices, and what's-new updates relevant to AI administration

The Entra AI Administrator role is the recommended RBAC role for Copilot management, separating day-to-day Copilot and agent administration from the broader privileges of Entra Global Admin. FSI organizations should assign this role to designated Copilot administrators to enforce least-privilege access and reduce the operational risk associated with over-privileged accounts.

Copilot Surface Coverage

Surface	Coverage	Notes
Microsoft 365 Copilot Chat	Full	User access, web search, agents, and Cloud Policy dependencies
Word / Excel / PowerPoint / Outlook	Full	Covered through Copilot access decisions plus workload-specific controls where applicable
Teams	Partial	Copilot rollout aligns with Teams meeting and messaging policies
Copilot Pages / Notebooks	Full	Governed through Cloud Policy plus SharePoint/Purview controls
Agents	Full	Managed through the Agents control plane and related agent settings

Governance Levels

Baseline

Assign Copilot administration to designated AI Administrators
Document current Copilot settings across Copilot > Overview, Copilot > Settings, Agents, and Cloud Policy
Review self-service purchase and PAYG status before broad rollout
Require documented approval before enabling Copilot for new populations or feature sets
Maintain a current inventory of agent access, Copilot Pages policy scope, and tenant-level Copilot availability

Implement segregation of duties so approval and implementation are not performed by the same person
Use group-scoped access for Copilot, agents, PAYG, and Copilot Pages policy decisions
Review Copilot settings monthly and compare against the approved baseline
Monitor Copilot and agent configuration changes through audit searches and governance reporting
Include Cloud Policy and billing policy decisions in the same change register as Copilot settings

Regulated

Require dual approval (technology + compliance) for new Copilot features, agent availability changes, or metered billing enablement
Restrict Copilot administration with PIM or equivalent time-bound privilege activation
Preserve configuration evidence and approval records for examination-ready retention periods
Include Copilot administration in SOX ITGC or equivalent control testing where relevant
Maintain a documented exception register for any setting that deviates from the approved baseline

Setup & Configuration

Step 1: Review Copilot Overview

Navigate to M365 Admin Center > Copilot > Overview and capture:

readiness and adoption indicators
recommended actions
security links to Purview and related controls
evidence of which governance team members review this dashboard

Step 2: Review Copilot Settings

Navigate to M365 Admin Center > Copilot > Settings and review the current settings model:

User access
Data access
Copilot actions
Other settings

Document which settings are enabled, who approved them, and which user groups are affected.

Step 3: Review Agent Governance Settings

Navigate to M365 Admin Center > Agents > Settings and confirm:

allowed agent types
sharing configuration
user access scope
any templates or defaults applied to agent publication workflows

Review the Agent 365 consolidated registry for cross-platform agent visibility, including Copilot Studio and third-party agents. Document any agents visible in Agent 365 that are not covered by existing governance policies.

Step 4: Review Copilot Security Pivot

Navigate to M365 Admin Center > Copilot > Security and review:

whether data access scope policies have been created to limit Copilot grounding sources
which users or groups are subject to data access restrictions
whether third-party model provider access is enabled or restricted for agents
document approved and restricted model providers alongside data residency implications

Step 5: Review Copilot Pages / Notebooks Policy

Navigate to https://config.office.com > Customization > Policy Management and document:

whether Create and view Copilot Pages and Copilot Notebooks is enabled
whether code previews are enabled
which users or groups the policy targets

Step 6: Review Billing and Self-Service Controls

Review:

Settings > Org settings > Self-service trials and purchases
Billing > Pay-as-you-go services
Cost Management

Document whether self-service purchases are blocked and whether any PAYG billing policies are active.

Step 7: Review Baseline Security Mode

Navigate to M365 Admin Center > Settings > Org settings > Security & privacy and confirm how Baseline Security Mode is being used as a complementary foundation for Copilot governance.

Financial Sector Considerations

Broker-dealers: Changes to Copilot access, web search, agents, or Pages should be evaluated alongside supervision, communication review, and books-and-records obligations before broad rollout.

Banking institutions: Copilot administration should be reflected in enterprise technology governance and change control records, especially where agent access or metered billing introduces new operational risk.

SOX-reporting entities: If Copilot is used in finance, reporting, or control-support workflows, configuration evidence and approval records should be retained in a form that supports internal and external audit review.

Copilot in admin centers: Microsoft now allows Copilot usage within the M365, Exchange, SharePoint, and Teams admin centers themselves. For FSI environments, this introduces a privilege escalation risk — an administrator using Copilot within an admin center to execute configuration changes represents a new threat surface where AI-assisted actions could modify security settings, user access, or compliance configurations. Organizations should evaluate whether to disable Copilot in admin centers for non-privileged admin roles, require approval workflows for Copilot-assisted admin changes, and ensure that all Copilot-assisted admin actions are captured in the unified audit log.

Privacy-sensitive environments: Read-only review roles such as Global Reader help governance teams collect evidence without granting unnecessary ability to change settings.

Advisory: Scheduled Prompts Governance

GA Feature — Requires Policy Assessment

Scheduled Prompts allow users to automate recurring Copilot interactions across Teams, Outlook, and Microsoft 365 Copilot Chat. Scheduled prompts require a Power Automate Standard license and are controlled via the optional connected experiences toggle in the M365 Admin Center.

FSI governance implications:

Automated communications: Scheduled prompts that auto-generate client communications, reports, or summaries may create FINRA Rule 2210 compliance obligations if the output is distributed externally. Organizations should assess whether scheduled prompt outputs are subject to the same supervisory review as manually drafted communications (see Control 3.5 — FINRA 2210).
Books-and-records: Auto-generated Copilot outputs from scheduled prompts may constitute business records under FINRA Rule 4511(a) and SEC Rule 17a-4. Organizations should verify that scheduled prompt outputs are captured in existing retention policies (see Control 3.2 — Data Retention).
Power Automate dependency: Scheduled prompts rely on Power Automate infrastructure. Organizations should coordinate with their Power Platform governance framework to ensure flow execution, error handling, and data access controls are applied to Copilot-triggered automations.
Access control: Administrators can control scheduled prompt availability via the optional connected experiences toggle in M365 Admin Center > Settings > Org settings > Services > Microsoft 365 on the web. FSI organizations should evaluate whether to restrict this capability to specific user groups during initial rollout.
Audit trail: Scheduled prompt executions should appear in Purview audit logs. Organizations should verify that audit log retention covers automated Copilot interactions and that scheduled prompts are distinguishable from interactive sessions in audit data.

Recommended actions:

Evaluate whether to disable scheduled prompts by default and enable for approved user groups
Update supervisory review procedures to address auto-generated Copilot outputs
Verify retention policy coverage for scheduled prompt outputs
Coordinate with Power Platform governance for flow-level controls
Include scheduled prompts in Copilot usage analytics monitoring (Control 4.5)

Verification Criteria

#	Verification Step	Expected Result
1	Access Copilot overview	Dashboard accessible to designated admins and reviewers
2	Review Copilot settings tabs	Settings documented across User access, Data access, Copilot actions, and Other settings
3	Verify role assignments	AI Administrator used as primary admin role; broader roles limited appropriately
4	Review agent settings	Allowed types, sharing, and user access reflect approved policy; Agent 365 registry reviewed for cross-platform agents
5	Review Copilot security pivot	Data access scope policies and third-party model provider settings are documented and approved
6	Review Cloud Policy scope	Copilot Pages / Notebooks policy matches intended user population
7	Review billing controls	Self-service purchases and PAYG status are documented and approved
8	Search audit logs for recent changes	Changes have corresponding approvals and evidence
9	Confirm baseline review cadence	Governance team has a documented review frequency