Microsoft plans to overhaul how external AI meeting assistants join Teams calls, introducing a new policy in 2026 that will automatically route these bots to the meeting lobby. Organizers will see a labeled notification and must explicitly approve entry, marking one of the most significant tightening moves against unsanctioned AI listeners in enterprise meetings.

A preview of the feature, outlined in internal Microsoft communications, describes a dedicated admin control called “External AI Meeting Bot Policy.” Administrators will be able to set the behavior to block all such bots, allow them only after organizer approval, or permit them unconditionally. The default is expected to shift from open access to lobby-and-approval mode, directly impacting third-party services like Otter.ai, Fireflies.ai, and countless other meeting transcription and summarization tools that join as guests.

The change arrives alongside a broader push by Microsoft to position Teams as the most secure collaboration hub for regulated industries. By funneling external AI bots through a transparent approval workflow, the company aims to quell growing concerns about data leakage, compliance violations, and even industrial espionage via unvetted third-party clients.

The Mechanics of the Lobby Control

According to the leaked details, when an external AI bot attempts to join a Teams meeting, it will no longer be silently admitted—even if anonymous or guest access is already enabled. Instead, the bot will be placed in the lobby with a distinctive label such as “AI Meeting Assistant – Pending Approval.”

The meeting organizer receives a prompt in the participant pane, highlighting the bot’s name and the fact that it is an AI tool requesting access. The organizer can then approve or deny entry. If approved, the bot joins like any other participant. If denied, it remains in the lobby until the meeting ends, at which point it is disconnected.

This system is separate from existing guest access and lobby policies. While Teams already allows admins to control lobby behavior for external participants, the new policy specifically targets entities that identify themselves as bots or automated assistants. Microsoft appears to be using the bot identity signal from Azure Active Directory (now Microsoft Entra ID) and the Teams client platform to distinguish AI-powered tools from regular human guests.

Administrator Configuration

IT admins will manage the policy via the Teams admin center under Meetings > Meeting policies > External AI meeting bot control. Three settings will be available:

  • Block all external AI meeting bots – Bots cannot join, no lobby prompt appears.
  • Allow after organizer approval – Bots go to the lobby; organizer must approve each one.
  • Allow all – Bots join directly without restriction (current behavior for most tenants).

The policy can be assigned globally or scoped to specific users or groups, giving granular control over which meeting organizers are protected. A corresponding PowerShell cmdlet is also in development for automated deployments.

Microsoft is not disabling the existing “Allow anonymous users to join meetings” or “Guest access” settings, but the new AI bot policy overrides them when a joining entity is classified as an external bot. This layered approach ensures that human participants are not affected, while AI assistants face additional scrutiny.

Why Now: The Rise of Unmanaged AI Scribes

The explosive growth of generative AI has flooded enterprise meetings with third-party note-takers. Tools like Otter.ai, Fireflies, and even built-in assistants such as Google’s Duet AI for Meet can record, transcribe, and summarize conversations—often without the meeting organizer’s explicit consent. In many cases, employees install these bots without IT approval, leading to shadow IT risks.

A 2024 survey by a leading compliance vendor found that 34% of organizations had at least one unapproved meeting bot active in a given week. Regulatory bodies in finance and healthcare have begun issuing fines when AI assistants inadvertently capture sensitive information without proper notice.

Microsoft’s own Copilot for Teams operates under the tenant’s compliance boundary, but external services do not. They may store transcripts on servers outside the organization’s control, violating data residency and retention policies. The new lobby control aims to put the decision back into the hands of meeting organizers, who can assess whether a bot is authorized before it starts recording.

Impact on Third-Party Bot Providers

For developers of AI meeting assistants, the policy represents a significant gate. Companies like Otter.ai and Fireflies have built businesses on frictionless joining—often requiring just a calendar invite to automatically send a bot to a meeting.

Under the new regime, their bots will be stuck in the lobby unless the organizer manually approves them each time. While this may seem cumbersome, Microsoft is likely to offer a path for third-party bots to gain “trusted” status through a certification program, similar to how some bots are approved for the Teams app store.

Insiders suggest that Microsoft will also release APIs allowing approved bots to be recognized and fast-tracked, provided the tenant admin has explicitly allowed them. This would preserve the user experience for widely trusted tools while maintaining security boundaries.

Community and Enterprise Reactions

Early feedback from IT administrators has been mixed. On one hand, security professionals applaud the move as long overdue. “Shadow AI tools are a nightmare for DLP and compliance,” said one regulated industry admin in a private preview group. “Having the organizer see exactly what’s trying to listen is a massive improvement.”

On the other hand, productivity advocates worry that the extra step will discourage the use of legitimate AI assistants that many teams rely on. “If I have to approve the bot for every single meeting, I’ll just turn it off,” a power user commented in an early adopter forum. “Microsoft needs to provide a whitelist.”

Indeed, the leaked documentation hints at a future “trusted external AI providers” list in the admin center, but that feature may not arrive until late 2026 or 2027. In the meantime, the lobby-and-approve model will be mandatory for upgraded tenants.

How It Compares to Competitors

Zoom and Google Meet have taken different approaches. Zoom’s administrative controls for third-party bots are relatively coarse—either allow all or block all—while Meet relies on Workspace admin policies that require explicit installation of add-ons. Microsoft’s lobby model adds a real-time human decision element that neither competitor currently offers.

This could give Teams a competitive edge in high-security industries, where every guest must be scrutinized. It also aligns with Microsoft’s broader zero-trust philosophy, which insists that no device or user should be trusted by default, even if they come through a seemingly legitimate channel.

Preparing for the Rollout

Microsoft has not publicly announced an exact GA date, but internal roadmaps point to public preview in Q2 2026 with general availability by mid-year. Tenants on the standard release channel will get the new default behavior automatically; those on targeted release may see it earlier.

Admins should start auditing which external bots are currently joining their Teams meetings. Teams reports in the admin center can show participant join records, including bot identities. Communication campaigns will be necessary to educate meeting organizers about the new approval process.

End-user training should emphasize that legitimate AI note-takers will still work but will require one-click approval. Organizations that wish to maintain seamless access for specific bots should begin planning to enroll those vendors in any upcoming trusted-partner program.

The Broader Security Picture

This policy is one piece of a larger security overhaul Microsoft is conducting for Teams. Recent updates have already introduced watermarking of shared content, end-to-end encryption for scheduled meetings, and restrictions on the use of Copilot in meetings with certain classifications.

The AI bot lobby control closes a gap that many CISOs have been highlighting: the ability of automated agents to join meetings without any visible trace. Now, the organizer becomes a gatekeeper, and the bot’s presence is no longer silent.

Microsoft is also expected to tie the policy into Microsoft Defender for Cloud Apps, allowing security teams to see real-time analytics of AI bot join attempts and block risky patterns automatically.

A Sign of Things to Come

As AI agents become more autonomous, the line between helpful assistant and intrusive listener blurs. Microsoft’s lobby approach may set a precedent for how we control all third-party agents in collaborative spaces. Already, there are discussions about extending similar policies to AI avatars and digital twins that represent remote colleagues.

For now, the immediate impact will be felt by millions of Teams users who have grown accustomed to bots silently joining calls. Some will welcome the transparency; others will see it as an annoyance. What is clear is that Microsoft is betting that security-sensitive organizations will prefer the extra step over the risk of unvetted AI listeners.

The company’s own Copilot will continue to work within the tenant’s trust boundary, giving it a native advantage over external tools. Critics might argue this is a lock-in tactic, but Microsoft maintains that the policy applies equally to all external bots, regardless of vendor, and is driven solely by customer security demands.

Actionable Takeaways

For IT departments, the check-list is straightforward:

  • Audit current external bot usage via Teams admin reports.
  • Communicate the upcoming change to all meeting organizers.
  • Evaluate which third-party AI assistants are business-critical and prepare to whitelist them.
  • Test the lobby experience once public preview is available.
  • Update your security policy to define acceptable AI meeting bot use.

For vendors of AI meeting assistants, now is the time to work with Microsoft on certification and to educate customers on how their bots will continue to work under the new model. A proactive approach will differentiate them from less-prepared competitors.

The lobby is about to become the new battleground for AI meeting security. Microsoft is drawing a clear line: no AI enters without a human’s permission.