Connect with us

NEWS

Teams Will Keep Invisible Meeting Archives for Copilot

Microsoft Teams will file invisible AI meeting archives by default in October 2026 so Copilot can answer after 120-day transcripts expire.

Published

on

Microsoft Teams will start writing AI meeting archive files in October 2026 so Copilot and Facilitator can keep answering questions after the transcript is gone. The files are on by default, they live in a tenant SharePoint Embedded container, and staff cannot open them.

That is the product. It is not a new Recap tab, and it is not a folder of notes people can browse. It is a long-lived memory layer for Microsoft’s meeting AI, built to outlast the transcript clock most tenants already use.

Copilot Will Remember Meetings After Transcripts Die

Microsoft 365 Roadmap item 558933 details state the point in plain language. Admins get a tenant-wide policy that lets Copilot and Facilitator hold key insights from meetings after transcripts expire, so those tools can still answer questions about older meetings. Admins also set how long that AI-generated archive is stored.

Admins can enable a tenant-wide policy that allows Copilot (and Facilitator) to retain key insights from meetings beyond transcript retention periods. This enables Copilot to continue answering questions about past meetings even after transcripts expire.

Microsoft 365 Roadmap, item 558933

The Microsoft 365 Message Center item MC1429018, last updated August 27, 2026, puts a calendar under that idea. Targeted Release begins in early October 2026 and should finish by mid-October 2026. Worldwide general availability begins in mid-October 2026 and should finish by late October 2026. Teams admin center controls show up about two weeks before archive file generation begins, and those October dates are the dates the controls arrive.

Roadmap status on August 28, 2026 was still In development, with a GA date of October CY2026. Microsoft Learn’s page on the feature, last updated July 26, 2026, still says AI archives are in preview for organizations in the Teams TAP program. TAP tenants can already be living with the files. Everyone else is waiting on the October drop.

MC1429018 is tagged as a major change with admin impact and user impact. People who feel it are Teams administrators and users who have Microsoft 365 Copilot or Facilitator turned on. The services named are Microsoft Teams, Microsoft 365 Copilot, and SharePoint Embedded.

A File Staff Cannot Open

Microsoft’s documentation on AI archives for Teams meetings describes a.meeting file, not a readable packet of quotes. The file holds meeting metadata such as title and date, plus a summary of key insights and discussion. No raw user data and no transcript excerpts go in. All content is AI-generated.

Users cannot open or view the archive. Copilot, Facilitator, and other Microsoft AI tools are the only consumers. If those tools use an archive to answer a question, the reply includes a citation that points at the meeting recap or the meeting invite, not at the.meeting file itself.

The files are indexed for search. Opening one from a search result does not preview the AI summary. It redirects to the recap page or the invite. The archive cannot be moved, renamed, or modified. That is the design, not a missing button.

Access follows the meeting, not the file. People invited to the meeting, and people later given access by sharing the recap, can get AI answers grounded in that archive. AI tools do not surface insights to anyone who lacks meeting access. Archive knowledge also stays inside the same tenant as the meeting organizer, so a guest from another company does not inherit the memory.

Deleting the human-facing record does not always delete the AI one. The meeting organizer can remove the archive by opening the recap, using the more menu, and choosing Delete recap content. If someone only deletes or expires the transcript or the recording, the AI archive stays. A Teams admin can also delete it in PowerShell with Remove-AIGeneratedKnowledge, which needs a ThreadId and can take an optional CallId.

If the organizer’s account is later deactivated, the archive still sits in a Teams-owned container until the tenant’s expiration time, or until a Purview retention policy says otherwise. When that clock runs out, the file is deleted and cannot be recovered.

Where Teams Stores the Meeting Archives

The files do not land in a user’s OneDrive. They sit in a Teams-owned SharePoint Embedded container, using Microsoft 365 File Storage Containers. SharePoint Embedded is an API-only store with no end-user site of its own. The first archive created for a tenant creates a new container. Later files reuse a container that still has space, or they open a new one. Those containers hold archive files only.

In a multi-geo tenant, the file is stored in the preferred data location of the meeting organizer. Admins cannot manage these containers through SharePoint Embedded APIs. Teams owns the lifetime. That is a different model from Loop, Copilot Pages, and Copilot Notebooks, which live in user-owned containers that show up under the Loop app name in the SharePoint admin center.

Purview still reaches the store. eDiscovery can search all SharePoint Embedded content through the All SharePoint Sites workload, or it can target a specific archive container. A Teams admin can look up that container URL in PowerShell with Get-AIGeneratedKnowledgeContainer, again using ThreadId and optional CallId. Holds and exports can run on the archive content. Retention policies for SharePoint classic and communication sites already include these files, so a separate location picker is not required unless the tenant wants a different rule via include or exclude on those container URLs. If a Purview retention policy and the Teams expiration setting disagree, Purview wins.

WHAT WE KNOW

  • Tenant store: Archives live in Teams-owned SharePoint Embedded containers, not in user OneDrive, and staff cannot browse them as files.
  • Search behavior: The files are indexed, but opening a hit sends the person to the recap or the invite rather than to the AI summary.
  • Same-tenant rule: Archive knowledge is shared only with users in the same tenant as the meeting organizer.

WHAT IS UNCONFIRMED

  • Copilot timing: Message Center says Facilitator uses the archive at release and Copilot will use it “in the future,” with no date attached.
  • Storage bill: SharePoint Embedded as a platform can be metered in Azure, but Microsoft has not said whether these Teams-owned archive containers add a separate charge.
  • Preview vs GA: Learn still limits the feature to the Teams TAP program, while Message Center schedules worldwide GA in October 2026.

That last gap is why compliance teams cannot treat a TAP tenant as a dress rehearsal for every other cloud. The October controls are the moment the rest of the fleet has to pick a policy.

Five Years Versus 120 Days

Here is the clock that makes the roadmap line real. Microsoft’s meeting recording policy says recordings and transcripts expire after 120 days by default. The AI archive default is five years (1,825 days). Microsoft even tells admins that longer archive expiration helps Facilitator and Copilot.

Those two clocks are not a pair. One is a storage cleanup setting for.mp4 files and transcripts in OneDrive or a channel library. The other is a knowledge setting for an AI file people never see. Expire the first and the second can still answer questions about that meeting.

TWO CLOCKS FOR THE SAME MEETING

Store Default life Who can open it If the transcript is deleted
Recording and transcript 120 days People with access in OneDrive or SharePoint The recording and transcript are gone
AI.meeting archive Five years (1,825 days) Nobody; Copilot and Facilitator only The archive remains unless recap content is deleted
Copilot only during the meeting No saved transcript, no archive Live answers only, then discarded Nothing remains to delete

Admins can change the archive expiration in the Teams admin center. They should weigh the five-year default for archive files against whatever transcript policy the tenant already runs. A company that burns recordings after 120 days and then leaves archives on for 1,825 days has not reduced meeting memory. It has moved that memory into a file nobody on staff can inspect.

Encrypted Meetings Stay Out, Barriers Do Not Apply

Not every call becomes a.meeting file. Generation needs all three of these at once: a transcript is created and saved, at least one Copilot-licensed participant is invited, and Allow Copilot and Facilitator is set to During and after the meeting. Miss any one of those and there is no archive.

Microsoft also withholds the file from three classes of meeting that already treat content as locked down.

WHEN TEAMS WILL NOT WRITE AN ARCHIVE

  • Encrypted labels: Meetings with sensitivity labels that include encryption do not get a.meeting file.
  • End-to-end encryption: E2E meetings are skipped, matching the existing promise that those sessions do not leave a transcript trail for AI.
  • Copy and forward limits: Meetings that restrict copying or forwarding also skip archive generation.

The gap on the other side is Information Barriers. Meeting archive files do not yet support them. A tenant that uses barriers to stop two groups from seeing each other’s files does not get that same wall on this store. eDiscovery and retention still apply. The barrier feature does not.

The practical off switch for a single sensitive meeting is the same control that already governs live AI. Set Allow Copilot and Facilitator to Only during the meeting and Teams will not save a transcript, so it will not write an archive. Set it to Off and Copilot, Facilitator, recording, and transcription all go dark for that meeting. Organizers who want a live assistant without a five-year shadow already have that path. They have to choose it before the meeting, not after the file exists.

How Admins Turn the Archive Off

No immediate action is required, Message Center says, which is another way of saying the default will do the work unless someone changes it. The setting path is Meeting Policies, then AI Memory and Archive, then Allow AI to archive meetings with.meeting file generation. Archive generation is on by default for each tenant. It can be turned off there. It cannot be turned on for only a subset of users.

That all-or-nothing rule is easy to miss. A finance group that wants Copilot memory and a legal group that does not cannot split the feature by policy assignment. The tenant is in or it is out. Per-meeting options remain the only finer cut, and those depend on organizers actually changing Allow Copilot and Facilitator.

Microsoft’s recommended list is short. Review the AI archive setting before rollout. Evaluate the default expiration of five years (1,825 days) and align it with compliance policy. Tell helpdesk and compliance stakeholders. Watch the Teams admin center, because file generation starts 2 weeks after the setting is released to the organization. Turning the feature off does not break existing Teams meeting tools. Microsoft calls archive generation an optimization for Copilot and Facilitator quality, including answers beyond transcript retention, depending on the expiration date the tenant sets.

THE PATH TO THE OCTOBER SWITCH

  1. April 8, 2026: Roadmap item 558933 is published as an AI-generated meeting archive for knowledge retention.
  2. July 15, 2026: Message Center item MC1429018 announces default-on.meeting file generation.
  3. July 26, 2026: Microsoft Learn describes the feature as TAP preview and documents SharePoint Embedded storage, deletion, and Purview hooks.
  4. August 27, 2026: MC1429018 updates the timeline to October 2026 and points admins at the Learn article.
  5. Early October 2026: Targeted Release admin controls begin to appear, with completion due by mid-October 2026.
  6. Mid-October 2026: Worldwide general availability begins, with completion due by late October 2026, and archive generation follows 2 weeks after each tenant receives the setting.

The two-week lag is the only built-in pause. Once the toggle is in the admin center, the tenant has that window to disable the policy or shorten 1,825 days before files start to accumulate. After generation begins, each new qualifying meeting writes a file that organizers cannot rename and users cannot open.

Facilitator Will Use the Archive at Launch

Facilitator is the first consumer. Facilitator in Teams meetings already sits in the room as a shared agent, taking notes, tracking the agenda, and answering in meeting chat where everyone can see the thread. Copilot in the same meeting stays private to the person who asked. Message Center says Facilitator will use the new archive to answer questions about meetings, with citations back to recaps or invites. Copilot gets that corpus later.

That order matters for anyone who thinks of Copilot as the only AI in the call. Facilitator is the agent that writes into a shared Loop notes surface and talks in the open chat. Grounding that agent on a five-year, user-invisible file means the shared assistant can keep a longer memory than the transcript the tenant thought it was deleting. Copilot’s later arrival extends the same memory into private prompts.

Microsoft’s privacy case is the missing raw transcript. The compliance case is everything else: default on, 1,825 days, no human preview, no Information Barriers, and a file that survives transcript deletion. Admins who want less of that memory have a tenant switch and a per-meeting AI option. Admins who want more of it are being told to lengthen the clock. The October controls are when that choice stops being theoretical.

Harry edits WinAddons, an independent news site that he owns and runs, covering Windows, Xbox, Azure, Microsoft 365, Teams, OneDrive, Outlook, the software built around them and Microsoft's business. His method comes from ten years in journalism, a reporter's years followed by an editor's, and the bulk of that decade has been spent watching Microsoft ship. His reporting starts with what Microsoft publishes: release notes and KB articles read in full, build numbers checked on an installed machine, MSRC advisories and the CVE records behind them, the Azure status history, lifecycle pages, store listings in the market they apply to, and the earnings releases and filings that carry the company's numbers. Every figure is checked against its source before publication, and a public corrections policy explains how mistakes are fixed and labelled. On security stories he does not publish exploit details before a fix is available, reporting what is affected and what to do instead. Pre-release features are labelled by channel and build, and a rumour is called a rumour. Readers can reach Harry at support@winaddons.com.

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending