How to Zoom Turn Off AI Companion (and Why There's No Single Switch)

If you want to zoom turn off AI companion, you need to know upfront: there is no master kill switch. Each feature has its own toggle, buried in different parts of the Zoom web portal. The AI Companion panel on your home screen, the in-meeting summary, the smart recording, the calendar sidebar, the Notes button: all separate settings, all defaulting to on. This guide walks through every one of them, explains who actually has the power to flip each toggle, and covers what happens when the toggle is grayed out and you can't touch it at all.

Key Takeaways

• There is no single opt-out for Zoom AI Companion. You must disable meeting summaries, smart recordings, in-meeting questions, the AI panel, and calendar/Notes features individually.
• Only the meeting host (or their account admin) controls whether AI Companion runs during a meeting. Participants cannot disable it themselves, only request that the host do so.
• If a toggle appears grayed out, your organization's admin has locked the setting. Your only options are to ask the admin or decline to host meetings on that account.
• Zoom routes some AI Companion features through third-party model providers, meaning meeting content may leave Zoom's infrastructure for processing.
• Some institutions (Stanford, among others) have disabled AI Companion site-wide by default. If yours hasn't, you're likely opted in already.

Why Is Zoom AI Companion On by Default?

Because adoption is the product goal. AI Companion usage jumped 68% quarter-over-quarter and quadrupled year-over-year among enterprise users. Those numbers don't happen by shipping features off and hoping people find them. Zoom made AI Companion available by default for millions of users, which means the feature is running unless someone actively goes in and kills it.

This is a pattern we see across the industry. Recording, transcription, and AI-assist tools ship "on" and bury consent rather than leading with it. The rationale is always the same: friction reduces adoption. From a product standpoint, it works. From a privacy standpoint, it means a therapist, a lawyer, or a job candidate joining someone else's Zoom call has no visibility into whether AI Companion is processing their words until a notification pops up mid-meeting.

A May 2026 report described the AI Companion panel on the Zoom home screen as functioning essentially as advertising for the paid AI add-on, with the company making it deliberately difficult to remove. That's the context. You're not just disabling a feature. You're opting out of a growth funnel.

How Do I Disable AI Companion for Meetings I Host?

You disable it feature by feature in the Zoom web portal, not in the desktop app. Sign in at zoom.us, go to Settings, and search for "AI Companion." You will find separate toggles for at least these features:

• Meeting Summary (automatically generates a summary after the meeting ends)
• Smart Recording (AI-generated chapters, highlights, action items from cloud recordings)
• In-Meeting Questions (lets participants ask AI Companion questions about what's been discussed)
• AI Companion for Meetings (the broad toggle that enables the sparkle icon in the toolbar)

Turn each one off. Then check that there isn't a separate toggle for "AI Companion" under your account-level settings versus your personal settings; as the Zoom Community discussion makes clear, there is no single option to opt out completely, and each feature has to be addressed individually.

This is the part that frustrates people. You think you've turned it off, then a sparkle icon reappears in your next meeting because you missed one toggle three screens deep. Check all of them.

How Do I Remove the AI Companion Panel from the Home Screen?

The AI Companion panel that appears on the Zoom desktop client home screen (and sometimes in the daily view) is a separate nuisance from the in-meeting features. CompanionLink's walkthrough covers the specific steps, but the short version: look in your Zoom desktop settings under "Home" or "General" and disable the AI Companion panel there. Note that Zoom has not provided a way to hide the calendar or Notes button as of this writing. Those persist regardless of your preferences.

If an uninvited bot appears in your meeting (showing up as a named participant), it may not be Zoom's AI Companion at all. Third-party tools like Otter or read.ai can join meetings as bots. Check Account Settings, then Apps, for anything you didn't install, and disable "Auto-join all meetings" on any app you find there.

What If the Toggle Is Grayed Out?

It means your organization's admin has locked the setting. You cannot override this as an individual user. Your options are:

1. Ask your IT admin to disable AI Companion at the account level or to remove the lock so you can toggle it yourself.
2. Refuse to host meetings on that account (host from a personal account instead, if your workflow allows it).
3. Accept it and move on.

None of those are great. But that's the architecture. Zoom's admin controls are hierarchical: account-level settings override group-level settings, which override user-level settings. If the admin locked it on, you're stuck.

How Do Admins Disable AI Companion for an Entire Organization?

Admins can permanently disable AI Companion account-wide through the Zoom Web Portal under Account Settings, then AI Companion. Toggle each AI Companion feature off, then click the lock icon next to each toggle. Locking the toggle prevents any user in the organization from re-enabling it for their own meetings.

This is the closest thing to a master switch that exists. It's still not one click. You still have to lock each sub-feature individually. But once locked, it stays off for all future meetings and all users on the account.

Some institutions have already done this at scale. Stanford disabled the AI Companion feature that uses in-meeting chats to generate summaries site-wide by default, specifically to protect sensitive information. Hosts on Stanford's Zoom account are no longer allowed to enable it.

What Happens If I'm a Participant, Not the Host?

You have almost no control. Only the meeting host controls AI Companion for meetings they host. As a participant, you cannot toggle it. You can request that the host turn it off: click the AI Companion icon if it's visible in the toolbar, and you can send a request to the host to disable meeting summary and meeting questions during the meeting. That's it. The host can ignore the request.

Features aren't active until the host manually enables them via the sparkle icon. When they do, other participants see a notification and a glowing icon. Zoom's own privacy resource page confirms participants can send disable requests, but the host makes the final call.

What About the New Consent Pop-ups?

Some institutions now require explicit acknowledgment. Starting January 26, 2026, when a host enables AI Companion features like meeting summary or smart recording, all participants must acknowledge and agree to an AI Companion Policy Disclaimer to remain active in the meeting. If you take no action, you can stay in the meeting, but your microphone and camera are disabled until you agree.

This is better than silent activation. It's still not great. "Agree or sit there muted with your camera off" is consent in the same way that "agree to the terms of service or don't use the product" is consent. It's technically a choice. Practically, if you're in a work meeting, you're going to click Agree.

Who Can See What AI Companion Records?

Zoom's privacy policy states AI Companion data is processed to deliver features, retained for a period, and can be shared internally within an organization's account. That means your meeting summary might be visible to your manager, your admin, or anyone with account-level access, depending on how your organization configured sharing.

Zoom currently routes some AI Companion features through third-party model providers, meaning user data may be transmitted to outside companies for processing. Zoom has stated it will not use user data to train AI models without consent, but trust was damaged when the initial launch did use user data for training, and the subsequent policy revision hasn't fully repaired that for many users.

This is worth understanding concretely. Your meeting audio is being transcribed. That transcription is being sent to a model provider for summarization. The summary is stored and shared within your org. Every step in that chain is a point where your words exist as data on someone else's infrastructure.

Why "Zero Retention" Claims Deserve Skepticism

We run an AI product (Selina), so we have direct experience with how LLM inference pipelines work. Here is the thing almost nobody in this industry says plainly: if your product sends data to a frontier model provider at inference time, some portion of that data exists on third-party infrastructure during processing. The retention window might be short. The provider might contractually agree not to train on it. But "zero retention" as an absolute claim is almost always false for any product that uses hosted LLM inference.

Zoom is not unique here. This is the structural reality of every product that bolts AI onto an existing platform and routes requests to external model providers. The question isn't whether your data touches third-party infrastructure. It does. The question is how long it persists, who can access it, and whether you were told before it happened.

When we built Selina, we made a deliberate choice: memory is encrypted at rest, but it is not end-to-end encrypted, because a slice of each request reaches a frontier provider at inference. We state that limit openly rather than pretending it doesn't exist. Operational metadata has a short retention window, not zero retention. These distinctions matter, and most products gloss over them.

What If I'm a Freelancer or Contractor Joining Someone Else's Meeting?

This is the most underexamined problem with Zoom AI Companion's architecture. If you're a freelancer, a therapist, a lawyer, or a job candidate joining a meeting hosted on someone else's Zoom account, you have zero control over AI Companion settings. The host's admin controls apply. You can't audit the settings before the meeting starts. You might not even know AI Companion is active until the notification appears after the host enables it mid-call.

For sensitive conversations (legal consultations, therapy sessions, medical discussions, salary negotiations), this creates genuine exposure. The person with the least power in the meeting has the least visibility into what's being recorded and processed.

Your practical options:

• Ask the host before the meeting whether AI Companion is enabled.
• Request they disable it if the conversation involves sensitive material.
• If they decline, decide whether you're comfortable proceeding. You can leave.
• For recurring sensitive conversations, consider hosting from your own account where you control the settings.

None of these are frictionless. That's the point. The friction is by design, just pointed in the wrong direction: it's easy to turn AI features on, hard to turn them off, and nearly impossible for a guest participant to do anything about it.

Does Disabling AI Companion Actually Stop All AI Processing?

It stops the features you toggled off. It does not necessarily remove every AI-adjacent feature from your Zoom experience. The AI Companion branding, the panel on the home screen, the calendar integration, the Notes button: these have separate settings and some cannot be hidden at all. You can disable the functional features (summaries, transcription, smart recording) while the UI chrome persists.

This is the fragmented opt-out problem. Compare it to how a privacy-respecting product should work: one switch, fully off, no residual UI, no ghost features. When you disable something, it should disappear. The fact that Zoom requires you to hunt through multiple settings pages, across both the web portal and the desktop client, and still leaves artifacts in the interface, tells you something about the incentive structure. They want you to see AI Companion. They want you to turn it back on.

Step-by-Step Summary: The Full Disable Checklist

For hosts and individual users:

1. Sign in to zoom.us (web portal, not the app).
2. Go to Settings.
3. Search "AI Companion."
4. Disable Meeting Summary.
5. Disable Smart Recording.
6. Disable In-Meeting Questions.
7. Disable AI Companion for Meetings (the broad toggle).
8. Check Account Settings, then Apps. Remove any third-party bots (Otter, read.ai) you didn't install. Disable "Auto-join all meetings" on any remaining apps.
9. In the Zoom desktop client, look for settings to remove the AI Companion panel from the Home screen.
10. Accept that the calendar and Notes buttons may not be removable.

For admins:

1. Sign in to zoom.us as an admin.
2. Go to Account Settings, then AI Companion.
3. Disable each AI Companion sub-feature.
4. Click the lock icon next to each toggle to prevent users from re-enabling.
5. Communicate the change to your organization so hosts know the sparkle icon is gone.

For participants on someone else's meeting:

1. Watch for the AI Companion notification when the meeting starts.
2. Click the AI Companion icon (if visible) to send a disable request to the host.
3. If the host declines, decide whether to continue or leave.
4. For future sensitive meetings, host from your own account.

Why Does This Matter Beyond Zoom?

Zoom is the most visible example of a broader pattern: AI features shipped on by default, with opt-out buried behind multiple settings, and participants given less control than hosts. This is the default posture of most platforms bolting AI onto existing products. The product ships, the AI is on, and consent is treated as a speed bump rather than a prerequisite.

The alternative is building with AI off by default, consent at the first interaction, and a single switch that means what it says. That's a design choice, not a technical limitation. The technology to give users a one-click disable exists. The incentive to ship it doesn't, because every toggled-off user is a lost data point and a lost upsell opportunity.

If you want an AI assistant that remembers your context across conversations without defaulting you into features you didn't ask for, start a free 7-day trial of Selina, no card required.

Michael C.

Founder & Principal Engineer, Selina Labs

Michael builds Selina, a privacy-first AI that remembers you across conversations. He ships security-sensitive AI in production — real attacks, real fixes, measured in minutes and dollars — and writes about privacy, security, and LLMs from that seat. Top Rated Plus and expert-verified on Upwork.

• www.upwork.com
• x.com
• medium.com
• selinalabs.io