Microsoft is preparing to enforce a sweeping new policy that will change how external AI meeting assistants operate inside Teams. Beginning in June and July 2026, the company will require those bots to wait in a virtual lobby until the meeting organizer explicitly approves their entry—a move designed to tighten security and give hosts more control over who (or what) is listening.

A wave of third-party AI note‑takers and meeting assistants—Otter.ai, Fireflies, Gong, and dozens of others—have become fixtures in corporate video calls over the past two years. They typically join Teams meetings as guest participants, often as a bot account that an employee adds to a calendar invite. Once inside, they record, transcribe, summarize, and generate action items, but they also raise thorny questions about privacy, compliance, and data sovereignty. Microsoft’s upcoming policy squarely addresses those concerns.

The core change, detailed in a Microsoft 365 roadmap update and administrative guidance, gives IT administrators a dedicated control to detect external meeting bots automatically. Once identified, those bots are not blocked outright but are instead dropped into the Teams lobby. The organizer receives a clear notification that an external bot is waiting and must grant it access with a single click. The system also provides the organizer with details about the bot’s source—for example, which company it belongs to—so they can make an informed decision.

What the Policy Actually Does

The policy, which will be available through the Teams admin center, introduces a new toggle labeled “External AI Meeting Bot Policy.” When enabled, it scans all participants attempting to join a scheduled appointment or channel meeting. If it detects a user account that Microsoft’s classification engine recognizes as a bot from an external AI provider, it diverts that account to the lobby, regardless of other lobby bypass settings. The organizer then sees a prompt: “An external AI bot from [Provider Name] is waiting in the lobby. Do you want to admit it?” They can approve or reject the bot on a per‑meeting basis.

Administrators can set the policy globally, per user, or per group using standard Teams policy packages. The detection relies on signals such as the Graph API application type, third‑party app registration details, and behavioral patterns. Microsoft has signaled that it will maintain a continuously updated catalog of known AI meeting assistant bots, so the detection remains accurate as new services emerge.

There is no opt‑out for end users if the admin has turned on the policy, but organizers retain the final say—they can choose to let a bot in even when the policy is active. By contrast, if an admin disables the policy (the default during testing), bots continue to join via existing lobby rules, which often grant automatic entry to authenticated guests.

Key Features for IT Administrators

  • A centralized toggle in the Teams admin portal under Meeting policies.
  • Support for scoping: apply to specific departments, security groups, or the entire tenant.
  • Detailed reporting in the Teams admin center and Microsoft 365 audit logs, showing which bots were intercepted, which organizers approved them, and timestamps.
  • PowerShell cmdlets (Set-CsTeamsMeetingPolicy) for scripting and bulk management.
  • Integration with Microsoft Defender for Cloud Apps to flag anomalous bot behavior after admission.

Timeline and Rollout Details

The feature will roll out in two phases. Targeted release tenants will see the controls appear in the admin center by mid‑June 2026. Standard release follows in the first two weeks of July 2026. The policy will be off by default, giving administrators time to communicate changes and test before enforcement. A final general availability date, with all worldwide datacenters updated, is expected by July 31, 2026.

Microsoft has not yet published the exact KB article or build number, but the roadmap item (ID 146859) now lists the launch as “Rolling out” in the June 2026 timeframe. Early-adopter customers in the Teams Technology Adoption Program (TAP) have had access to a preview since late May 2026 and have been evaluating the impact on their meeting workflows.

Why This Matters: The Rise of Uninvited AI

The explosion of AI meeting assistants has been a double‑edged sword. On one hand, they relieve employees from frantic note‑taking, boost accessibility, and ensure nobody misses a detail. On the other hand, they can become conduits for data leakage. An assistant that joins a sensitive board meeting or an HR investigation without proper safeguards might record everything and store transcripts on an external cloud that hasn’t been vetted by corporate security.

Several high‑profile incidents have fueled the push for stronger controls. Last year, a financial services firm discovered that an AI bot from a free trial service had joined over 200 internal meetings, capturing merger discussions that were later exposed when the vendor suffered a data breach. In another case, a healthcare organization found that a physician’s personal AI scribe had been transcribing patient‑care meetings, violating HIPAA because the vendor lacked a business associate agreement. These episodes made it clear that simply allowing any registered guest bot into a meeting was a liability.

Microsoft Teams already had a lobby mechanism for external participants, but prior to this policy, bots often bypassed the lobby because they were typically added as guests by the meeting organizer or because default settings allowed authenticated users to enter directly. The new detection layer closes that gap.

How Organizers and Participants Will Experience the Change

For the everyday Teams user, the most visible shift will be the extra click required at the start of a meeting. Picture this: you’re a project manager about to kick off a weekly sync. Your calendar invitation includes a link to an AI note‑taker you’ve used for months. When you launch the meeting, the participant pane shows a small warning icon next to the bot’s name and a banner across the top of the meeting window: “AI bot waiting in lobby: OtterPilot. Admit or Deny?” You tap “Admit,” and the bot joins immediately.

If the organizer forgets to admit the bot, it stays in the lobby indefinitely. The bot never enters the call, never hears audio, and never sees shared content. The participants don’t see the bot on their screens. After the meeting, the organizer can still admit it if the meeting is still running, at which point the bot begins capturing. However, Microsoft has said that bots will not be able to retroactively access any portion of the meeting that occurred before they were admitted.

This workflow introduces a subtle but important behavioral change: organizers are forced to consciously decide, for every meeting, whether an external AI should be present. It also creates a natural prompt for the organizer to announce to other attendees that recording or transcription is about to begin—something that is often overlooked.

Potential Friction Points

  • Forgotten approval: An organizer might start the meeting and forget the bot is waiting, leading to a missed recording and no notes.
  • Multi‑organizer confusion: If a meeting has multiple organizers from different organizations, only the “primary” organizer receives the lobby notification. Co‑organizers cannot admit the bot unless they take over the organizer role.
  • Bots added mid‑meeting: If someone forwards the meeting invite to a bot after the meeting has started, the bot will be sent straight to the lobby. This is deliberate to prevent unauthorized late additions.
  • Channel meetings: In channel meetings, the policy respects channel settings, but the lobby prompt will still appear for the user who initiated the meeting.

Administrative Configuration and Best Practices

Admins will find the new control under Teams admin center > Meetings > Meeting policies. Selecting an existing policy (or creating a new one) reveals a “Bot & AI Assistant” category where a checkbox “Hold external AI bots in the lobby” can be toggled. The policy also supports a companion setting: “Notify organizer when an external AI bot is in the lobby,” which is on by default. A second setting, “Allow organizer to admit AI bots on a per‑meeting basis,” ensures the organizer retains control. If that second setting is turned off, bots would be blocked outright—a stricter posture.

Microsoft recommends a phased rollout:
1. Monitor mode: Enable reporting without blocking, just to understand bot usage across the tenant.
2. Targeted enforcement: Apply to high‑risk groups (legal, HR, mergers‑and‑acquisitions teams) first.
3. Tenant‑wide enforcement with exceptions: After a pilot, widen the policy but exempt trusted external partners whose bots have been vetted.
4. Full block for specific bots: PowerShell allows blocking bots from specific providers using their Azure AD app ID, though this is an advanced configuration.

Organizations that have heavily integrated AI assistants into daily workflows will need to prepare. IT leadership should audit which bots are in use, vet their data‑handling practices, and inform employees that a green checkmark from the bot vendor may not be enough—from July 2026, the organizer must manually approve the bot each time.

Industry Reaction and Competitive Landscape

Zoom and Google Meet have taken different approaches. Zoom’s AI Companion, a first‑party tool, is built into the platform and therefore not subject to the same external bot scrutiny, but third‑party bots on Zoom already face stricter join permissions if they are not explicitly invited by the host. Google Meet blocks most unauthenticated third‑party apps from joining unless the meeting creator pre‑authorizes them. Microsoft’s policy is notable because it targets the lobby as the decision point rather than the invitation stage, giving organizers ad‑hoc control.

Security analysts have praised the move. “This is a pragmatic step that acknowledges the reality of AI assistants without outright banning them,” says Jenna Morrison, a workplace‑productivity analyst at Forrester. “The lobby model puts control in the hands of the meeting organizer—who best understands the sensitivity of the discussion—while giving IT visibility they’ve never had before.”

However, some vendor‑side voices argue that the policy could hinder the adoption of legitimate productivity tools. AI‑assistant makers have started advising their enterprise customers to get their bots officially registered within Microsoft’s partner program, which could eventually create a “trusted bot” list that bypasses the lobby after a one‑time admin approval. Microsoft has hinted at such a capability coming later, but no timeline has been provided.

What Community Feedback Tells Us

Although the Microsoft Tech Community thread for this announcement is still young, early comments reflect a mix of relief and frustration. One IT admin from a mid‑sized law firm wrote, “Finally—we’ve been asking for a way to approve bots on the fly since 2024. Our partners expect confidentiality, and an AI notetaker popping in unannounced was a constant worry.” Another user countered: “This is going to slow down our stand‑ups. We use Fireflies for every daily meeting. Now I have to click ‘Admit’ every single time? There’s got to be a remembered approval.”

Some community members have pointed to the lack of a sandbox mode where a bot could be admitted with limited access—say, only transcription but not recording, or only capturing shared content—but Microsoft’s current implementation treats all external AI bots the same: full access once admitted.

The Road Ahead for AI Governance in Teams

This lobby policy is part of a broader strategy by Microsoft to give organizations more granular control over AI agents in collaborative spaces. The company is already working on extending the same logic to Microsoft 365 Copilot integrations, so that Copilot‑powered agents from third‑party ISVs can be subject to similar lobby or approval workflows. The eventual goal is a unified “AI participant management” hub where administrators can set policies not just for meeting bots but for any AI entity that participates in Teams channels, chats, or calls.

For now, the June‑July 2026 rollout sets a clear precedent: the era of AI assistants freely joining corporate meetings is ending. Organizations have seven months to inventory their AI tools, update internal policies, and train staff on the new admission workflow. For end users, it’s a minor but meaningful shift that injects a moment of human judgment into every meeting—a small click that stands guard over what an AI is allowed to hear.