Microsoft did not ask for permission. Starting May 20, 2024, Power BI Copilot default activation went live across all qualifying tenants meaning organisations that had not previously enabled Copilot found it switched on automatically. A second wave followed in September 2025, enabling the standalone Copilot experience by default for every tenant where the core Copilot setting was already active. For IT directors and Fabric tenant administrators, this represents a governance decision that has already been made on their behalf and one that requires a deliberate response rather than passive acceptance.

What Changed: The Shift to Default-On Activation

Prior to May 2024, Copilot in Power BI required explicit administrator action to enable. Tenant admins had to locate the Users can use Copilot and other features powered by Azure OpenAI setting in the Microsoft Fabric admin portal and toggle it on, along with a cross-geo data processing setting for organisations outside the United States and France. The friction of that activation process combined with unresolved concerns about data retention, meant that a large proportion of qualifying tenants had not enabled Copilot at all.

Microsoft addressed both the friction and the data concerns simultaneously. The Power BI Copilot default activation change eliminated the need for manual enablement by switching the feature on for all tenants that met the licensing and capacity requirements. Simultaneously, Microsoft resolved the two most significant blockers that had caused administrators to withhold consent: prompt data retention and cross-geo data processing for EU tenants. The practical consequence is that any organisation with a qualifying Fabric or Power BI Premium capacity that had not proactively disabled Copilot before May 20, 2024 will have had it switched on automatically.

"Default-on activation transfers the governance decision from opt-in to opt-out. Organisations that have not explicitly evaluated the Copilot readiness of their semantic models should do so now, not after users have already begun querying data through an AI interface."

The Data Privacy Update That Made Default Activation Possible

The most significant obstacle to Copilot adoption in regulated enterprise environments was Microsoft's previous policy of retaining user prompts for abuse monitoring purposes for up to 30 days. For organisations subject to data governance frameworks financial services, healthcare, legal, government the prospect of employee queries containing client names, deal values, patient indicators, or commercially sensitive terms being retained on Microsoft infrastructure was incompatible with their data handling obligations.

Microsoft resolved this in full. As of the May 2024 update, no user prompt data and no customer data is stored by Microsoft for any purpose, including abuse monitoring. The abuse monitoring approach was redesigned to function without data retention. This change removes the primary compliance blocker that had prevented many enterprise organisations from enabling Power BI Copilot and is the prerequisite condition that made default activation defensible from a governance standpoint.

For data protection officers and legal teams reviewing Copilot readiness, this policy change should be formally documented as part of the organisation's AI tool assessment record. The absence of prompt retention does not eliminate all data governance considerations, the semantic models that Copilot queries still contain data that must be governed through row-level security and workspace access controls but it removes the most acute risk that the earlier architecture presented.

EU Data Boundary: What Changed for European Tenants

Prior to the May 2024 changes, organisations with Fabric tenants located within the EU data boundary faced an additional barrier: using Copilot required explicitly enabling a cross-geo data processing setting that permitted query data to be processed outside the EU. For organisations with strict data residency requirements particularly those subject to GDPR-mandated data localisation obligations this was a categorical blocker regardless of the prompt retention policy.

The May 2024 update resolved this by updating the geo-mapping logic. Organisations with capacity regions inside the EU data boundary no longer need to enable the cross-geo setting. Copilot queries from EU tenants are now processed within the EU by default, with no data leaving the boundary. This change makes Power BI Copilot default activation meaningfully different in its risk profile for European organisations compared to the pre-May 2024 architecture and removes what was, for many EU-based enterprises, the last structural barrier to activation.

Licensing and Capacity Requirements

The capacity constraint is the most common point of failure for organisations expecting Copilot to simply appear after default activation. The Copilot button may be visible in the Power BI interface while the underlying capacity does not meet the F64 minimum a scenario that causes confusion without a clear error message. Tenant administrators should verify that all workspaces where Copilot use is intended are assigned to a qualifying capacity before communicating availability to end users.

The Enabled vs Disabled Decision Framework

The decision to leave Power BI Copilot default activation in place or to proactively disable it is not primarily a technical question, it is a data readiness and governance question. Copilot's quality of output is directly proportional to the quality, consistency, and descriptiveness of the underlying semantic models it queries. An organisation with well-governed semantic models, clearly named measures, complete relationship definitions, and documented field descriptions will see materially better Copilot results than one where models were built for technical rather than conversational use.

Arguments for Keeping Copilot Enabled

If your organisation's Power BI semantic models are mature measures are clearly named, business logic is encapsulated in certified datasets, row-level security is applied correctly, and workspaces are governed by role-based access then leaving Copilot active delivers immediate productivity value. Analysts can interrogate data through natural language without requiring DAX knowledge. Executives can ask questions directly against reports without needing to interpret visual layouts. The productivity gain is genuine and measurable, particularly for organisations where report consumers currently rely on analysts to answer ad hoc data questions.

Arguments for Disabling Until Ready

If your semantic models contain ambiguous measure names, undocumented business logic, inconsistent relationship structures, or uncertified datasets that mix governed and ungoverned data, enabling Copilot in that environment creates risk. Copilot may confidently return inaccurate answers derived from poorly defined models and users have no reliable way to distinguish a correct Copilot response from an incorrect one without querying the underlying data themselves. The result is eroded trust in analytics output, which is considerably harder to rebuild than it was to establish. Disabling Copilot until data model governance meets a defined readiness standard is the more defensible choice in this scenario.

How to Disable Copilot Before or After Default Activation

Disabling Copilot is a two-step process in the Microsoft Fabric admin portal. Navigate to Admin Portal → Tenant Settings → Copilot and Azure OpenAI Service. The primary setting Users can use Copilot and other features powered by Azure OpenAI, can be toggled off to disable Copilot entirely for all users in the tenant. Alternatively, it can be scoped to specific security groups if the intent is to run a controlled pilot rather than a full disable.

For organisations that want to acknowledge the default activation without fully enabling the feature thereby preventing a future automatic re-enable Microsoft's opt-out mechanism for the September 2025 standalone Copilot wave works differently: briefly toggling the standalone Copilot setting on and then immediately off signals to Microsoft's system that the administrator has made a deliberate choice, preventing automatic re-enablement at the next default-on wave.

The Standalone Copilot Experience: A Second Default-On Wave

The September 2025 default activation wave introduced a distinct additional layer: the standalone Copilot experience, also described as Chat with Your Data. This is a full-screen, chat-based AI interface accessible from the Power BI left navigation bar that allows users to query any report, semantic model, or Fabric data agent they have access to independently of a specific report page. It is a broader, more open-ended interaction surface than the in-report Copilot panel.

This second wave applied to all tenants where the core Copilot setting was already active. The standalone experience requires the tenant-level Copilot setting to be on capacity-level enablement alone is insufficient to activate it. For organisations that had enabled Copilot at the capacity level but not at the tenant level, the standalone experience will not activate automatically, and administrators should be aware of this distinction when auditing what did and did not change in September 2025.

Next Steps: Preparing Your Semantic Models for Copilot

Whether your organisation chooses to keep Power BI Copilot default activation in place or defer it, the preparatory work is the same: audit the semantic models that power your most-used reports and assess their readiness for conversational AI queries. This means reviewing measure naming conventions, adding field descriptions to ambiguous columns, certifying trusted datasets in the Power BI service, and confirming that row-level security is correctly applied so that Copilot cannot surface data that the querying user is not authorised to access.

Organisations that invest in this readiness work before enabling Copilot broadly will see adoption rates and user satisfaction significantly higher than those that enable the feature without preparation. Copilot is a multiplier on the quality of the data model beneath it — not a substitute for it. A well-governed Power BI governance programme is the foundation that makes Copilot trustworthy at enterprise scale.

If your organisation needs to assess Copilot readiness, design a semantic model governance framework, or evaluate whether your current Fabric capacity tier supports the AI features you intend to use, our Power BI consulting team at Numlytics can structure that assessment and remediation programme. We work with data and IT leadership teams across the US, UK, Australia, and UAE. Speak with a certified Power BI consultant to get started.

For a comprehensive view of Fabric capacity sizing including how F64 maps to Power BI Premium P1 and what workloads each tier supports, see our guide to choosing the right Microsoft Fabric SKU as an executive.