How to Disable Copilot for Employees in 2026: What the Windows Policies Remove, and What They Leave Running
.jpeg)
Short answer: On managed Windows 11 devices, IT can hide Copilot with the TurnOffWindowsCopilot policy and uninstall the free Microsoft Copilot app with the newer RemoveMicrosoftCopilotApp policy, but both only touch the app on the device. Our take: neither policy stops an employee from opening Copilot, ChatGPT or any other AI in a browser with a personal account, so the control that actually sticks is tenant-aware enforcement on the endpoint, which is what dope.security applies across every browser and app.
"How do I disable Copilot?" is one of the most searched Windows questions of the year, and on a personal PC the answer is a settings toggle. On a company fleet, it's a policy question with a catch. Most organizations don't want Copilot gone. They want the consumer version gone and the Microsoft 365 Copilot they're paying for, the one covered by enterprise data protection, kept. For the security side of that licensed version, our guide to Microsoft Copilot security risks and controls is the companion read.
This guide covers the Windows policies that exist today, exactly what each one does according to Microsoft's documentation, and the gap they leave. If your real goal is controlling what data employees share with Copilot, read our piece on how to control the data employees share with Copilot and Gemini alongside this one.
Can you disable Copilot for all employees?
Partly. You can remove or hide the Copilot app experience on managed Windows devices with policy. You cannot use Windows policy to stop someone from visiting a Copilot website in a browser, signing in with a personal Microsoft account, or switching to a different AI tool entirely. That second half requires a web-layer control.
It helps to separate the two products Microsoft ships under similar names. Microsoft Copilot is the free, consumer-oriented app and web experience. Microsoft 365 Copilot is the licensed, work-account experience tied to your Entra ID tenant. Most security teams want the first one gone and the second one available. Windows policy can only partly express that distinction, and it can't express it at all inside a browser.
Option 1: TurnOffWindowsCopilot (the old switch)
The TurnOffWindowsCopilot policy is the setting most older guides point to. Per Microsoft's Policy CSP documentation, enabling it means users can't use Copilot and the Copilot icon won't appear on the taskbar.
There's an important note on the same page: Microsoft says the TurnOffWindowsCopilot policy isn't for the new Copilot experience that rolled out to Windows 11 and Windows 10 devices. In other words, it was written for the original sidebar-style Copilot in Windows. If your fleet is on the current app-based Copilot, this policy alone won't get you where you want to go.
Option 2: RemoveMicrosoftCopilotApp (the new uninstall policy)
Microsoft added a more targeted policy under the WindowsAI area of the Policy CSP: RemoveMicrosoftCopilotApp. Microsoft describes it as a way to uninstall Microsoft Copilot from devices in a targeted way. Coverage from Petri and The Register reported it reaching broad availability in 2026 through both Group Policy and MDM tools like Intune.
The fine print matters. According to Microsoft's documentation, the policy only applies when all of these are true:
- Both apps are installed: Microsoft 365 Copilot and Microsoft Copilot are both present on the device.
- The user didn't install it: the Microsoft Copilot app wasn't installed by the user.
- It's not in active use: the Microsoft Copilot app wasn't launched in the last 28 days.
Microsoft's edition table lists Enterprise, Education and IoT Enterprise, and not Pro. Secondary coverage also reports that the uninstall is a one-time action and that users can reinstall the app afterward. And by design, it leaves the paid Microsoft 365 Copilot app in place, which is usually what you want.
So RemoveMicrosoftCopilotApp is a clean-up tool for idle consumer Copilot installs on Enterprise and Education fleets. It's not a block. An employee who uses the consumer app every day, installed it themselves, or runs Windows Pro is outside its scope.
Option 3: App control and browser policy
To stop the app from coming back, teams layer application control on top. Microsoft's IT-pro guidance for the app-based Copilot points to AppLocker for blocking the app package, and Microsoft Edge has its own policies for the Copilot entry points inside the browser.
These are worth deploying. But notice the pattern: each control covers one surface. The taskbar. The installed app. One browser's sidebar. Employees have more surfaces than that, and AI vendors keep adding new ones.
The gap: Copilot is also a website
Here's where the "disable Copilot" project usually stalls. Even with every Windows and Edge policy set, an employee can open Chrome, go to the Copilot website, and sign in with a personal Microsoft account. Or skip Copilot entirely and use personal ChatGPT, Claude or Gemini. Windows policy doesn't see any of that, because it's just HTTPS traffic in a browser.
That's the same lesson security teams learned with ChatGPT: blocking the obvious entry point pushes usage somewhere less visible. Our analysis of why blocking ChatGPT doesn't work covers the pattern in detail. The goal isn't zero AI. It's AI on your terms: enterprise accounts allowed, personal accounts blocked, sensitive data kept out of both.
Doing that requires seeing the full HTTPS request, including which account is signing in, on every browser and desktop app. DNS filtering can't see accounts, and a browser extension only sees its own browser. It takes SSL inspection on the endpoint.
What about Microsoft 365 Copilot itself?
Most of the time, the licensed Microsoft 365 Copilot isn't the thing you want to disable. It runs under your Entra ID tenant, respects your existing Microsoft 365 permissions, and is the version your organization actually signed a contract for. The security work there is different: oversharing in SharePoint and OneDrive, sensitivity labels, and making sure Copilot can't surface files to people who shouldn't see them.
That's a data-at-rest problem, and it's worth solving before you widen Copilot access. dope.security's CASB Neural finds every externally shared Drive and OneDrive file, explains what makes each one sensitive, and lets you make it private again in one click. Clean up the exposure first, and the licensed Copilot has less to overshare.
The consumer side is the opposite problem. It sits outside your tenant, so none of your Microsoft 365 permissions, labels or retention policies apply to what an employee pastes into it. That's why the personal-account question matters more than the app icon.
Don't forget Macs and mobile
Windows policies only govern Windows. Microsoft ships Copilot experiences on macOS and on phones too, and every one of them is reachable in a browser. If part of your fleet runs on Macs, the RemoveMicrosoftCopilotApp policy simply doesn't exist for those devices.
A control that works at the web layer on the endpoint closes that gap on laptops. dope.security's agent runs natively on Mac and Windows, so the same tenant rules and DLP policies apply whether an employee opens Copilot on a ThinkPad or a MacBook. Personal phones on cellular data are outside any corporate web gateway, which is one more reason to keep sensitive files off devices you don't manage.
Windows policy vs endpoint AI governance
Here's how the native Windows controls compare with an endpoint approach like dope.security when the goal is "consumer Copilot off, work Copilot on."
- Taskbar and app: Windows policy hides the icon or uninstalls an idle consumer app; dope.security doesn't manage installed apps, it governs the traffic those apps and browsers send.
- Browser access: Windows policy doesn't cover a Copilot or ChatGPT website opened in Chrome or Firefox; dope.security inspects web traffic on the device across browsers.
- Personal vs work accounts: Windows policy can't tell a personal sign-in from a work sign-in on the web; dope.security's AI Analytics shows personal versus enterprise-licensed account usage, and Cloud Application Control restricts Microsoft sign-ins to your corporate tenant.
- Other AI tools: Windows policy is Copilot-specific; dope.security discovers ChatGPT, Claude, Gemini, Copilot and other shadow AI from one screen and lets you allow, warn or block by app.
- Sensitive data: Windows policy has no content inspection; Dopamine DLP classifies uploads and prompts by meaning, with coverage that includes Copilot, ChatGPT and Claude, and blocks sensitive content before it leaves.
- Edition limits: RemoveMicrosoftCopilotApp excludes Windows Pro; dope.security's agent runs on Windows and Mac regardless of edition.
The two layers work well together. Use Windows policy to tidy the device. Use endpoint AI governance to control what actually leaves it.
A practical rollout for IT teams
If you're building a "consumer Copilot off" plan this quarter, this sequence keeps productivity intact while closing the gaps.
Start with discovery. Before you remove anything, find out who uses which AI tools and with which accounts. dope.security's AI Analytics shows every AI app surfaced on your endpoints, who used it, and how much data moved, including personal versus enterprise account breakdowns. You'll often find Copilot is the smallest part of the picture.
Clean up the device. Deploy RemoveMicrosoftCopilotApp to eligible Enterprise and Education devices, set TurnOffWindowsCopilot where older builds remain, and add app control to stop reinstalls where your policy calls for it. Document the exceptions, especially Pro devices.
Govern the web layer. Allow your Microsoft 365 tenant and block personal Microsoft sign-ins with Cloud Application Control, then decide per app whether consumer AI tools are allowed, warned or blocked. Validate the exact Copilot sign-in flows in a short pilot, since Microsoft changes its consumer and work entry points often. Our guide to blocking personal ChatGPT while allowing ChatGPT Enterprise walks through the same tenant pattern.
Protect the data. Turn on Dopamine DLP so sensitive files and prompts are classified by meaning and blocked before upload, whichever AI tool the employee picked. For the Copilot-specific angle, see our Copilot DLP guide.
Deployment doesn't need to be a project of its own. A Fortune 100 company rolled dope.security from 900 to over 18,000 devices in a matter of weeks, deploying silently via Intune, as told in the Fortune 100 deployment story. That's the same Intune you're already using for the Copilot policies.
See how it fits your fleet: explore dope.security AI governance, the Fly Direct secure web gateway, or book a 20-minute demo.
The bottom line
Windows gives IT real tools to hide and uninstall consumer Copilot, and you should use them. But they stop at the edge of the app. The moment Copilot, or any other AI, opens in a browser with a personal account, only a control that reads the actual web request on the device can tell work from personal and keep sensitive data in. Tidy the desktop with policy, then govern the traffic with dope.security.
Frequently Asked Questions
How do I disable Copilot in Windows 11 for all users?
On managed devices, use Group Policy or MDM. TurnOffWindowsCopilot hides the original Copilot experience, and RemoveMicrosoftCopilotApp uninstalls the consumer Microsoft Copilot app on eligible Enterprise and Education devices. Neither policy blocks Copilot or other AI tools in a web browser, which requires a web-layer control such as dope.security.
Does RemoveMicrosoftCopilotApp remove Microsoft 365 Copilot?
No. The policy targets the consumer Microsoft Copilot app and only applies when Microsoft 365 Copilot is also installed. Microsoft's conditions also require that the user didn't install the app and hasn't launched it in the last 28 days.
Does the Copilot removal policy work on Windows 11 Pro?
Microsoft's edition table for RemoveMicrosoftCopilotApp lists Enterprise, Education and IoT Enterprise, not Pro. Organizations running Pro devices need another approach, such as application control combined with endpoint web governance.
Can employees still use Copilot in a browser after it's disabled?
Yes. Windows policies act on the app and taskbar, not on web traffic. An employee can open the Copilot website in any browser and sign in with a personal Microsoft account unless a web-layer control blocks personal sign-ins, which is what dope.security's Cloud Application Control is designed to do.
Should we block Copilot entirely?
Most organizations shouldn't. Blocking AI outright tends to push employees to less visible tools. A better pattern is to allow the licensed Microsoft 365 Copilot tied to your tenant, block personal accounts, and use DLP to keep sensitive data out of every AI tool.
How do I see who is using Copilot with a personal account?
You need visibility into the sign-in itself, which lives inside encrypted web traffic. dope.security's AI Analytics discovers AI apps on every endpoint and breaks usage down by personal versus enterprise-licensed accounts, so you can see the gap before you set policy.


.jpeg)
.jpeg)

