Why Turning Off AI in Collaboration Tools Matters
Disabling AI features in Microsoft Teams and Edge is the process of using built-in controls to stop AI notetakers, Copilot meeting assistants, and autonomous browser agents from recording, analysing, or acting on your business communications so that sensitive conversations stay private and enterprise collaboration security policies remain in charge of which AI tools are allowed to operate.
If you run IT for a company, or you handle sensitive projects, you have probably felt that uncomfortable moment when an AI bot joins a meeting and nobody is sure who invited it. Tools like Otter or Fireflies can appear as guests, record everything, and send data to their own clouds, which means information is leaving your tenant without anyone watching it go. The good news is that Microsoft is adding clear privacy controls: you can disable Teams AI features mid-meeting, block unknown external bots by default, and decide how far Copilot Cowork is allowed to go in Edge. This article walks you through a practical way to treat those controls as part of your normal security hygiene rather than an emergency response.
Controlling Meeting AI and External Bots in Microsoft Teams
Microsoft has split Teams AI controls into two big levers: what happens inside the meeting, and which external AI notetakers can get in at all. First, licensed organisers and presenters now get an in‑meeting toggle that turns Meeting AI on or off, covering Copilot, Facilitator, and meeting recap, and it can be switched at any time during a live meeting. That means you can keep AI on while you review project updates, then cut it off when the conversation moves into more sensitive topics, without ending or recreating the meeting. Second, at the tenant level, IT can set a default that blocks every identified external AI bot automatically, rather than relying on organisers to spot and reject them in the lobby.
Under the hood, external bot access is controlled with a policy setting called ExternalBotAccessMode in the Teams admin center. There are three values: AllowAllBots, RequireApprovalWhenDetected (the current default), and BlockDetectedBots, which turns bots away before they ever reach the lobby. This setting applies to external third‑party bots only; Microsoft 365 Copilot and bots registered as Entra ID applications inside your own tenant are handled through separate licensing and permissions. Policies can also be scoped per user group, so you can put legal, finance, HR, and executive teams behind a hard bot block while leaving a more flexible approval model for less sensitive departments. In practice, this gives you AI notetaker settings that match your real risk profile rather than a one‑size‑fits‑all switch.
Managing Copilot Cowork Privacy Controls in Microsoft Edge
On the browser side, Microsoft has added a new toggle to Copilot Cowork in Edge so users can decide how much autonomy they are comfortable with. In the Copilot and AI settings page, the option "Allow Cowork to take actions on your behalf" now governs whether the Cowork agent is allowed to automatically create tabs, take screenshots, or interact with web pages for you. This sits just below the "Browse with Copilot" menu, which has gone through several name changes—from Copilot Actions to Auto Browse before settling on its current label. Cowork itself runs in a hidden Edge tab and is designed to automate some web‑based tasks while you continue work in your main browser window.
For end users, the practical privacy choice is straightforward: if you are working with sensitive systems or signed‑in business apps, turning off the "Allow Cowork to take actions on your behalf" toggle means Copilot Cowork cannot autonomously interact with those pages. Microsoft emphasises that the agent only operates on websites where users are already signed in and requires the latest version of Edge. It is also still in preview and limited to Edge, not other browsers. From a Microsoft Copilot privacy controls perspective, this gives you a middle path: you can keep Copilot available for browsing help while denying it the right to click around or capture screenshots unless you are confident that fits your organisation’s policies.
A Simple Workflow to Keep AI Aligned With Enterprise Collaboration Security
To make these controls stick, treat them as one workflow that connects policy, meeting behaviour, and browser settings. Here is a practical sequence you can follow as an IT lead or privacy‑aware team owner.
- Decide which groups need strict protection (for example, legal, finance, HR, executive teams) and which can use AI more freely; document where sensitive conversations happen most often.
- In the Teams admin center, set ExternalBotAccessMode to BlockDetectedBots for your high‑risk groups, and leave RequireApprovalWhenDetected or AllowAllBots for lower‑risk groups where external AI notetakers are approved.
- Train organisers and presenters to use the in‑meeting toggle: keep Meeting AI on for routine discussions where insights and recap are useful, then disable it when the conversation moves into sensitive territory.
- Ask users of Edge with Microsoft 365 Copilot access to review the "Allow Cowork to take actions on your behalf" toggle and turn it off when working in signed‑in business apps or handling confidential data in the browser.
- Review your AI policies regularly, checking that tenant‑level settings, meeting behaviour, and Copilot browser controls all match your current enterprise collaboration security and compliance needs.
The main gotcha here is assuming one control is enough. Bot blocking does not replace app permission policies, sensitivity labels, recording controls, or clear guidance on which AI tools are approved; all of those still have a place in your governance stack. The meeting toggle respects existing tenant policies and compliance controls, so if those are loose, your in‑meeting decision will be too. Think of these features as extra brakes on a car that already needs good steering and speed limits.
Is a Privacy‑First AI Setup Worth the Effort?
Putting time into these controls is worth it if your organisation cares about where its data goes. Microsoft is tightening AI notetaker controls because the third‑party market has grown and governance problems have grown alongside it. When bots capture meeting audio and ship transcripts to their own services, it raises transparency and lawful basis questions, especially when participants were never told their words were being processed by a bot. At the same time, native Meeting AI is designed so organisers can capture insights when discussions are valuable and turn off AI when conversations shift to sensitive topics.
The takeaway: treat AI as something you switch on with intent, not something that quietly listens by default. Use the in‑meeting toggle to disable Teams AI features when you need privacy, block external bots that you have not explicitly approved, and set Copilot Cowork limits that match your risk appetite in Edge. Do that, and AI becomes a helpful colleague rather than an uninvited guest in your most important conversations.
