The feature hasn't shipped yet. Microsoft lists it as In development, so treat everything below as a planned change, not something you can turn on today.
What the roadmap entry says
Microsoft's description for item 571305 is short. It promises to stop blocking users from uploading files with MPIP sensitivity labels to Copilot Studio agents on the M365 channel. It says agents will process that content based on the user's or the agent's account permissions, with label protections "enforced end-to-end."
Here is what the listing contains:
| Field | Roadmap value |
|---|---|
| Roadmap ID | 571305 |
| Product | Microsoft Copilot Studio |
| Status | In development |
| Platform | Web |
| Release rings | Preview and General Availability |
| Cloud | Worldwide (Standard Multi-Tenant) |
| Preview estimate | October 2026 |
| GA estimate | October 2026 |
Preview and general availability both have the same month-level estimate of October 2026. That's ambitious for a feature still marked "In development." Microsoft's roadmap page also says its dates are estimates and that "all information is subject to change." Entries can be removed when a feature ships, is canceled or is postponed. Government clouds and sovereign tenants aren't listed. The entry only covers Worldwide multi-tenant.
Section summary: This is a planned capability with an October 2026 target. It covers file uploads to Copilot Studio agents on the Microsoft 365 channel, on the web, in the commercial Worldwide cloud.
Why labeled files have been a problem
Sensitivity labels exist to make sure protection travels with a document. That's the point of them. The trouble starts when an agent needs to read a protected file. If the agent ignores the label, you have a data leak. If it rejects every labeled file, the agent is useless for confidential work.
For many organizations, the second option has been a real problem. If finance labels every budget workbook "Confidential" with encryption, an agent that refuses labeled uploads can't help with any budget question. Users then copy the content into an unlabeled file to get around the block, which removes the protection the label was meant to provide.
The roadmap entry describes a middle path: accept the labeled file, but process it only within the permissions the label grants.
The detail admins need to know: VIEW alone isn't enough
Microsoft's existing documentation already explains how Copilot experiences handle encrypted, labeled content. Nothing official yet confirms the new upload scenario will use the same rules, but this is the most reliable guide available.
Microsoft's Copilot data protection architecture documentation says Copilot can only summarize or reference content that the user is authorized to access. When encryption is applied, the user must have EXTRACT and VIEW usage rights for Copilot to interact with the content. The Copilot Studio page on sensitivity labels in agent responses says the same about agents: for an encrypting label, users need EXTRACT as well as VIEW before the agent will return the data.
This affects how you set up labels. A user who can open and read a protected file in Word might still get nothing from an agent if the label doesn't grant EXTRACT. This is intentional, and admins can use it on purpose. Microsoft's Purview guidance says if you don't want Copilot and agents to summarize these files when they're at rest, use sensitivity labels that apply encryption without the EXTRACT usage right.
The same Microsoft documentation lists other cases where agents can't read protected content. We don't yet know whether these apply to the new upload feature, but they're worth checking now:
- User-defined permissions: Microsoft says user-defined sensitivity label permissions can block Copilot from extracting and interacting with the file content. For example, Copilot agents can't read files that have user-defined sensitivity label permissions.
- Double Key Encryption: Because Double Key Encryption (DKE) is intended for your most sensitive data that is subject to the strictest protection requirements, Copilot and agents can't access this data.
- Teams labels: Sensitivity labels that protect Teams meetings and chat aren't currently recognized by Copilot and agents.
Section summary: Being able to open a file is not the same as letting an agent read it. Check which labels grant EXTRACT before users start uploading files and filing helpdesk tickets.
How this fits with existing label support in Copilot Studio
Copilot Studio already handles labels for content it reaches through knowledge sources. Microsoft Learn lists the supported sources as SharePoint, OneDrive for Business, SQL, Dataverse, Cosmos, Azure Blob Storage, Word, Excel, Outlook, Office groups and Office users. That support is on by default for agents with supported knowledge sources. The documentation is still marked as preview.
When an agent uses labeled content, users see a shield icon showing the highest label applied to the material behind the answer, and each citation shows its own label. On the web the shield appears below the response. In Teams it appears above the response and shows label details when you hover over it.
The roadmap item covers a different path: a user uploading a file directly into a conversation, rather than the agent retrieving it from a connected source. Microsoft hasn't said whether uploaded files will show the same shield and citation labels. That would make sense, but it isn't confirmed.
What Microsoft hasn't said yet
The roadmap entry leaves a lot unanswered. Missing details include:
- Which file types are supported, and any size limits
- Licensing requirements
- Whether admins must enable anything at the tenant or environment level
- Which label configurations are supported, beyond what existing Copilot documentation suggests
- When the agent's identity is used instead of the user's. The entry mentions "user's/agents account permissions" but doesn't explain when each applies.
The last point matters most. If an agent ever processes a file with its own identity instead of the user's, the agent's permissions decide what it can reach. Security teams should expect detailed documentation on that before allowing the feature for high-sensitivity labels.
How to prepare
You can do this groundwork now, before the feature arrives:
- List your encrypting labels and note which ones grant EXTRACT to which groups. Those labels decide what an agent can actually use.
- Look for labels with user-defined permissions. Under current Microsoft documentation, agents can't read those files, which can confuse users.
- Decide which labels you want kept out of agents and remove EXTRACT from them on purpose, rather than leaving it to defaults.
- Check which of your agents are published to the Microsoft 365 channel. That's the only deployment the roadmap entry covers.
- Test in preview first. Use a non-production environment with a few labels at different protection levels, and confirm that blocked content stays blocked.
Bottom line
This is a useful change. Today, protected documents and Copilot Studio agents often don't work together, so users work around the protection. Microsoft's plan is to accept labeled files and limit processing to what the label permits, which fits how Purview already applies to Copilot.
The roadmap description is very brief, though, and the October 2026 target is only an estimate. Until Microsoft publishes full documentation, plan around rules you can check yourself, especially the VIEW-plus-EXTRACT requirement. For now, the main work for admins is cleaning up label permissions.