How do you stop AI notetaker bots from joining your Teams, Zoom and Meet calls?

The exact admin settings and per-meeting steps that keep Otter, Fireflies, Read.ai and similar bots out of Microsoft Teams, Zoom and Google Meet calls, why the bots get invited in the first place, the lawsuits and consent rules behind the worry, and what blocking breaks.

By Andrej Ruckij · · 17 min read

How do you stop AI notetaker bots from joining your Teams, Zoom and Meet calls?

By Andrej Ruckij · November 1, 2026 · updated November 4, 2026

TL;DR: In Microsoft Teams an admin can now block them outright with PowerShell: Set-CsTeamsMeetingPolicy -Identity Global -ExternalBotAccessMode BlockDetectedBots. That block mode rolled out between late August and late September 2026; the matching option under Teams admin center > Meetings > Meeting policies > “Manage external bots and their access to meetings” may appear there as the rollout completes. The default since this summer already holds detected bots in the lobby until someone admits them. Zoom and Google Meet have no bot detection that I could find, so there you lean on sign-in requirements, waiting rooms and removing the bot by hand. The bot usually arrives because someone on the invite connected Otter or Fireflies to their calendar, so the lasting fix is at that source, plus a written rule on which notetaker your company allows.

My read after going through the docs: Teams is now the only one of the three platforms with a real switch for this. Everywhere else you’re managing the door by hand, and even Teams says some bots won’t be detected. So treat the platform settings as the lock, and the calendar clean-up and a company rule as what actually keeps the problem small.

Why do AI bots join meetings in the first place?

Almost always through somebody’s calendar. A person signs up for Otter, Fireflies, Read.ai or a similar tool, connects their Google or Microsoft calendar, and from then on the tool sends its bot to every meeting on that calendar, including yours if they’re a guest on it.

The lawsuit against Otter describes exactly this. Justin Brewer’s sales call in February 2025 “was recorded because another participant on the call had OtterPilot running”, as tl;dv’s summary of the case puts it. The complaint, as NPR quoted it, says “an Otter Notetaker may join the meeting without obtaining the affirmative consent from any meeting participant”, and that the software typically asks the host for permission but not the other people on the call.

Case Western Reserve University, which blocked Fireflies.ai, Read.ai and Fathom across Zoom, Teams and Google Meet in September 2025, gave the same reason: the apps “often attempt to join future meetings by default—even when other attendees did not consent.”

That has a practical consequence. If the bot comes from your own staff’s calendars, you can stop it at the source. If it comes from a guest’s calendar, you can only stop it at the door.

How to do it

Who this is for: business owners and IT admins on Microsoft 365, Zoom or Google Workspace, and meeting organisers without admin rights (each platform has a section for you). Who it’s not for: if you need a meeting to be unrecordable, none of this gets you there. A guest can still run a notetaker app on their phone next to the speaker, which one frustrated Zoom user pointed out in the community thread I read.

How to prevent AI bots from joining Teams meetings (admin)

Needs a Teams administrator. About 15 minutes, and changes can take up to 24 hours to apply.

  1. Sign in to the Teams admin center, expand Meetings and select Meeting policies.
  2. Open the policy you want to change. Global (Org-wide default) covers everyone; a custom policy covers only the users or groups you assign it to.
  3. In the Meeting join & lobby section, find Manage external bots and their access to meetings.
  4. Pick one:
    • When detected, require approval before joining is the default. Detected bots wait in the lobby “regardless of the lobby setting of the meeting” until someone admits them.
    • Blocking is set in PowerShell (below) with BlockDetectedBots, which Microsoft’s cmdlet reference describes as: “When detected, external third-party meeting bots and meeting assistants are blocked from joining meetings hosted in the organization.”
    • Do not detect bots (AllowAllBots) switches detection off. Microsoft’s own warning: “They may get mistakenly admitted to meetings.”
  5. While you’re in the same section, set Who can admit from lobby to organizers and co-organizers. Microsoft’s admin guide recommends this, so a presenter can’t wave a bot in past the warning.
  6. Select Save.

PowerShell version, if you manage Teams that way:

Set-CsTeamsMeetingPolicy -Identity Global -ExternalBotAccessMode BlockDetectedBots

Microsoft’s admin page still shows examples with Set-CsTeamsEventsPolicy. The parameter is documented on Set-CsTeamsMeetingPolicy, so use that one.

How to check it worked: run Get-CsTeamsMeetingPolicy -Identity Global and look at ExternalBotAccessMode. Then, when a bot tries to join one of your meetings, it should either be refused or sit in the lobby labelled as a bot. UC Today describes the lobby section as “Suspected threats”.

Dates, so you know what you have. Detection and lobby labelling (Microsoft notice MC1251206, published 13 March 2026) rolled out June to August 2026 and shipped switched on. The block option (roadmap item 566201, notice MC1459141) rolled out from late August to late September 2026, according to Everywan’s write-up and the roadmap tracker. It ships switched off: it does nothing until you assign it. Microsoft’s admin page still lists only the two options above, so a block option may appear in the admin center as the rollout completes; the PowerShell value is already in the cmdlet reference.

What it does and doesn’t catch. Teams detects bots from “a combination of infrastructural and behavioral signals” during the join. Microsoft lists two limits on its own page: “Some external bots may not be detected”, and detection “may occasionally misclassify human participants as bots”. Microsoft 365 Copilot and apps registered in your own Entra ID are outside this setting, per Everywan’s write-up of the notice. Tony Redmond at Office 365 for IT Pros notes that vendors in Microsoft’s bot identification programme, including Otter.ai, Read AI and Recall.ai, show up as identified, while unknown ones are flagged as “Unregistered or system-identified bots”.

How do you block AI bots in Teams without the new setting?

The older controls still help, and they catch bots that detection misses. Same place: Teams admin center > Meetings.

  1. Turn off anonymous join if you don’t need it: Meetings > Meeting settings > Participants > Anonymous users can join a meeting > Off. Bots often join without signing in to a work account (one vendor’s help page, Avoma’s, says its bot joins without signing in to a Teams account), so this shuts them out, along with any human guest who isn’t signed in. You can do it per group instead with the Anonymous users can join a meeting meeting policy.
  2. Send outsiders to the lobby: set Who can bypass the lobby to People in my org (guests and external people wait) or Only organizers and co-organizers for sensitive groups.
  3. Add a CAPTCHA for people who aren’t signed in or come from untrusted organisations. This one is PowerShell only in what I read:
Set-CsTeamsMeetingPolicy -Identity Global -CaptchaVerificationForMeetingJoin AnonymousUsersAndUntrustedOrganizations

Organisers can change the lobby defaults in their own meeting options. If you need them locked, Microsoft points to meeting templates or sensitivity labels, which need Teams Premium.

How to disable an AI bot in a Teams meeting (no admin rights)

For any organiser. Two minutes per meeting.

  1. In the Teams calendar, open the meeting, select View event, then Options. During a meeting it’s the Controls menu > View more options.
  2. Set Who can bypass the lobby? to People in my org or Only organizers and co-organizers.
  3. Set Who can admit from the lobby to organizers and co-organizers.
  4. With Teams Premium, turn on Require unverified participants to verify their info before joining: they get a one-time passcode by email.
  5. In the meeting, don’t admit anything flagged as a bot or named like “Otter.ai” or “Fireflies.ai Notetaker”. If one is already in, remove it from the participant list and tell the guest why.

How to prevent AI bots from joining Zoom meetings

I found no Zoom setting that detects bots the way Teams now does. A Zoom Community thread on this, with 14,167 views when I read it, still had admins asking for one in September 2026, and the steps below are what admins and university IT teams use instead. The menu paths come from university help pages, since I couldn’t find Zoom’s own article on this.

Account-wide (Zoom admin, web portal). About 20 minutes.

  1. Settings > Meeting tab: turn on Only authenticated meeting participants and webinar attendees can join meetings and webinars. California College of the Arts’ guide says Otter.ai and other bots “are not able to login to Zoom accounts”, so they can’t join with this on.
  2. Same tab: turn on Waiting Room, then Edit Options and set Who should go in the waiting room? to Users not in your account.
  3. Same tab: turn on Block users in specific domains from joining meetings and webinars, click the pencil icon and add the bot domains you see, for example otter.ai and read.ai. The University of Illinois Springfield page is blunt that this “is not a complete list”.
  4. If you allow guests who aren’t signed in, turn on the option that makes guests who aren’t signed in solve a CAPTCHA (UIS lists it under the “Join from your browser” link setting in your profile settings).
  5. Remove notetaker apps your own staff installed: sign in at marketplace.zoom.us as admin, click Manage, go to Admin App Management > Apps on Account, pick the app (say Otter.ai), Manage App, and disable or remove it. That’s Zoom support’s answer as one admin quoted it in the community thread. Zoom’s own caveat: “External participants will still be able to utilize Otter.ai”. One admin who couldn’t find Otter there went to Admin App Management > Permissions and turned off app requests instead.

How to check it worked: join one of your meetings from a browser without signing in. You should hit the sign-in wall or the waiting room.

The limits, from people running this. The admin who started that thread in June 2025 had authentication on, Read.ai, Otter.ai and Fathom blocked as domains, and still had “to kick several of these AI notetakers” from every meeting. Another admin wrote in August 2026 that some notetakers “join using the meeting participant’s authenticated account rather than their own domain”, which gets past both the sign-in rule and the domain block. That fits a change Zoom made for developers on February 23, 2026: apps built on its Meeting SDK that join meetings outside their own Zoom account must now “attribute to a user”, either appearing as owned by a named person or joining as that user, instead of joining anonymously. My read: a bot riding on a guest’s own signed-in account is exactly what an authentication rule can’t stop.

Per meeting, as host:

  1. When scheduling, under Security, tick Require authentication to join and Waiting Room.
  2. Once everyone you expect is in: Security > Lock Meeting.
  3. To remove a bot: Participants, then More next to it, then Remove.
  4. If things get out of hand: Security > Suspend Participant Activities.

The “Stop Zoom AI features” button a Zoom community manager suggested in August 2026 turns off Zoom’s own AI Companion in that meeting. It doesn’t remove an outside bot.

How to keep AI notetakers out of Google Meet

Google’s docs don’t mention third-party bots at all. You control who gets in.

Admin (Google Workspace). About 10 minutes.

  1. In the Admin console go to Apps > Google Workspace > Google Meet > Meet safety settings.
  2. Under Domain, choose who can join meetings created by your organisation. Under Access, choose which meetings your users can join. The tightest option your business can live with is the one to pick; an all-external sales team can’t close it to the domain only.
  3. Turn on Warn for external participants, so outsiders, bots included, are marked External in the People list.
  4. Under Meet video settings, Supplemental add-ons lets you block add-ons, including featured third-party ones.

Per meeting, as host:

  1. Meet already makes people knock: anyone not on the calendar invite, and anyone not signed in to a Google account, must “request to join”. Read the name on each request before admitting it.
  2. Turn on Host management (it’s off by default): in Calendar, Video call options > Host controls, or in the meeting, Host controls at the bottom right.
  3. For sensitive calls, open the event in Calendar, go to the Waiting room tab and switch it on.
  4. To remove a bot: Show everyone, More actions next to it, Remove from the call, then Block so it can’t rejoin. The block only covers that meeting code.

How to stop it at the source

If the bot belongs to one of your own staff, platform settings are the slow way. The fast way is to disconnect the notetaker from their calendar. A Zoom user who fought Otter for months put it plainly: you “will need to remove it from your Calendar app. That’s what allows it to connect to all of the video calls.”

For the whole company, an admin can see and block every app connected to staff Google or Microsoft accounts, notetakers included. I wrote up those exact steps for Google Workspace and Microsoft Entra here: How do you remove third-party access from a Google account (and your company’s)?

If the bot belongs to a guest, you can’t reach their calendar. Ask them, or rely on the door controls above.

It depends on where the people are. US federal law needs the consent of only one party. The Reporters Committee’s recording guide lists about 11 states that need everyone’s consent: California, Delaware, Florida, Illinois, Maryland, Massachusetts, Michigan, Montana, New Hampshire, Pennsylvania and Washington. For calls across states, it advises assuming the stricter law applies.

The lawsuits are where this gets real:

  • Otter.ai. Four class actions were consolidated as In re Otter.AI Privacy Litigation (5:25-cv-06911, N.D. Cal., Judge Eumi K. Lee). They claim violations of the federal Wiretap Act and California’s Invasion of Privacy Act, which allows $5,000 per violation. In August 2026 the judge partly dismissed the case, but let the claims “predicated on real-time interception, third-party eavesdropping and collection of biometric voice data” go ahead, per Mealey’s. No class has been certified and nothing is proven.
  • Fireflies.ai. Two Illinois suits under the state’s biometric privacy law, Cruz (filed December 2025) and Fricker (March 2026), allege voiceprints were made without consent. Source: tl;dv, which sells a competing notetaker.
  • Granola, which tl;dv lists as a bot-free tool, was sued in July 2026 over capturing meetings “without most participants knowing”, also per tl;dv. So “no bot in the call” doesn’t mean no risk.

The best-known story behind all this predates the suits: in September 2024, engineer Alex Bilzerian said Otter kept transcribing after a Zoom call with investors ended, and emailed everyone on the invite a transcript of their private conversation afterwards. Attorney Joseph Lazzarotti, as UC Today reports, sees the host-consents-for-everyone model as legally shaky in all-party states like California and Illinois.

I didn’t research EU or UK rules for this piece, so I won’t guess at GDPR here. If your calls involve people in Europe, check that separately.

For your own company, a one-page policy covers most of it: which notetaker is approved, that it’s announced at the start of every recorded call, which meetings are never recorded (HR, legal, board), and that staff don’t connect unapproved tools to work calendars.

What does blocking AI bots break?

Four things, in the order I’d expect to hear complaints about them.

  • Your own sales and success team’s tools. Call recorders join Teams calls as outside bots. One vendor’s help page (Avoma) shows the fix vendors suggest: setting detection back to “Do not detect bots”. With blocking on, those calls lose their notes unless you give that team a separate policy.
  • Clients who send a bot instead of attending. One Zoom user was “furious” when a client sent only their notetaker to a presentation, then decided that removing it would just mean the client never got the information. If you block bots on external calls, you’ll have that conversation with customers.
  • Real people. Teams says detection can mislabel humans as bots. Organisers can admit them and mark “This is not a bot”, but someone has to be watching the lobby.
  • Accessibility. Some people rely on transcripts. Blocking the outside bot is easier to defend when you offer a built-in transcript instead.

Everywan, an IT consultancy, suggests a sensible rollout for Teams: watch the detection data for a few weeks, then block per group, starting with management, legal, HR and anyone holding personnel data, and tell staff who meet outside guests before you switch it on.

When should you allow a notetaker instead?

When the notes are worth having and the data stays inside a contract you control. Case Western’s answer to its own block was to point staff to “Zoom AI Companion” and “native captioning and transcripts”, tools covered by its existing contracts. On Microsoft 365, Tony Redmond’s view is that the problem is “confidential company information is leaving the tenant boundary”, and that some tenants without Copilot may reasonably allow outside bots. Zoom has AI Companion, and Google Workspace has a Google AI note-taking setting under Meet Gemini settings.

My read: for most small businesses the sensible setup is one approved notetaker (ideally the platform’s own), announced on every call, and block-or-lobby for everything else. Banning all AI notes tends to push people to phone apps you can’t see at all.

What I’d do this week

If you’re on Teams, check that your Global meeting policy still says “When detected, require approval before joining” and set Who can admit from lobby to organizers and co-organizers. Then make a custom policy with BlockDetectedBots for leadership, HR and legal. On Zoom, turn on authenticated-only join and a waiting room for users not in your account. On any platform, ask your team which notetakers are connected to their calendars, and write down which one you allow.

Sources

More of this every day on the channel: https://t.me/aidaily_business