Chrome Enterprise and Chromebooks

Block AI extensions in the Google Admin console

AI agent extensions can read every page a user opens and some can click, type and submit forms on their own. On managed Chrome browsers and Chromebooks, the Google Admin console can stop them, as long as you know which extension IDs to block.

Next step

AI Agent Blocklist: the Chrome extension IDs of AI agents, ready for the Google Admin console

Agent extensions, AI sidebars and browser automation add-ons, each with its tool name, category and risk grade, refreshed every day so new ones are covered.

From $99 a month for one organisation

See plans Ask a question

Why AI extensions are a problem on managed Chrome

An extension runs inside the user's signed-in browser. It sees what the user sees and can act with the user's sessions. When that extension is an AI agent, the browser becomes a place where an outside service reads and does things in your company accounts.

It reads every page

Many AI sidebars ask for access to all sites so they can summarise or answer about the open tab. That includes webmail, the CRM, HR systems and internal wikis.

It acts as the user

Agent extensions click buttons, fill forms and move between tabs. They do it inside sessions where the user is already signed in, so no extra login stands in the way.

Page content leaves the company

To answer a question, the extension sends page text, and often screenshots, to its AI service. You have no contract with that service and no say over retention.

Web pages can give it orders

Text hidden in a page, an email or a document can be read by the agent as an instruction. This is prompt injection, and it turns any website into a possible controller of the agent.

Updates change behaviour silently

Chrome updates extensions in the background. A tool that was harmless last month can ship new code, new permissions or a new owner without the user noticing.

Lookalikes copy trusted names

Fake AI sidebars copy the name, icon and interface of real ones. Users install the copy because it looks familiar and the store listing looks clean.

Dozens of extensions

In December 2024, researchers reported that fake store policy emails gave outsiders access to extension developer accounts. Altered updates then reached dozens of extensions with millions of users.

Lookalike sidebars

In December 2025, researchers reported extensions copying the look of an AI sidebar that sent users' AI chat conversations and open tab addresses to an outside server at fixed intervals.

Gartner's advice: in December 2025 Gartner published "Cybersecurity Must Block AI Browsers for Now". Agent extensions bring the same agent into a normal Chrome window, so the same reasoning applies to them. See our page on prompt injection in AI browsers for how the attacks work.

Signs that AI extensions are already in your fleet

Most admins find agent extensions only after a user asks why one stopped working. These checks show you what is installed before you change any policy.

  • Open the apps and extensions report for enrolled browsers in the Admin console and sort by number of installs. Check the current console for the exact report name.
  • Look for extensions whose names include AI, GPT, Copilot, agent, assistant, sidebar or autopilot. Then look up the ID, not the name.
  • Check extension requests from users. A rising number of requests for AI tools tells you demand is there and blocking needs a clear message.
  • On one test device, open chrome://management to see what the organisation manages, and chrome://policy to see which policies arrived.
  • Review requested permissions. Access to all sites, tabs, scripting, cookies, downloads or the debugger is common for agents.
  • Compare Chromebooks used by students with staff browsers. Student units often allow more than anyone intended.
  • Check DNS or firewall logs for AI agent domains. Extension traffic shows up there even when the console report lags.
  • Ask the help desk which AI tools users mention. Tickets about a sidebar or an agent tab are a good early signal.

Three ways to control AI extensions in Google Admin

The Admin console gives you three levers. Most organisations use a blocklist of agent extension IDs for staff, an allowlist for students, and permission limits as a safety net for everything else.

Lever 1Block by extension ID
  • Add each agent extension by its 32 letter ID.
  • Set the installation policy to Block.
  • Works with the mode Allow all apps, admin manages blocklist.
  • Needs a current list, because new agents appear weekly.
Lever 2Allowlist mode
  • Choose Block all apps, admin manages allowlist.
  • Optionally let users request extensions.
  • Strongest control, more requests to review.
  • Our list helps you answer requests quickly.
Lever 3Permissions and site access
  • Block risky permissions for all extensions.
  • Stop extensions from touching chosen sites.
  • Protects webmail, finance and admin portals.
  • Does not stop the extension elsewhere.
Not sure which model fits? Our guide on allowlist or blocklist for AI agents compares the work each one creates, and the Chrome extension allowlist page shows how to build an approved list.

Block AI extensions in the Google Admin console, step by step

These steps apply to managed ChromeOS devices and to Chrome browsers enrolled in Chrome Browser Cloud Management. Menu names change from time to time, so check the current admin console if a label differs.

  1. Create a test unit.

    Make a small organizational unit or group with one test user, one Chromebook and one enrolled browser. Policies on groups override organizational units, so check group membership too.

  2. Open the extension screen.

    Go to Devices > Chrome > Apps & extensions > Users & browsers and select the test unit on the left.

  3. Add the agent extension by ID.

    Use the add button and choose the option to add a Chrome app or extension by ID. Paste the ID from our file and confirm the tool name before you save.

  4. Set the installation policy to Block.

    The options are Allow install, Force install, Force install + pin and Block. Choose Block and save. Repeat for each ID, or script it for long lists.

  5. Pick the overall mode.

    In the additional settings for the Chrome Web Store, choose Allow all apps, admin manages blocklist for staff, or Block all apps, admin manages allowlist where you want an approved list only.

  6. Limit permissions and sites.

    In the additional app settings, open Permissions and URLs. Block permissions you never want extensions to have, and list sensitive sites that extensions must not interact with.

  7. Test, then widen.

    Sign in on the test devices, try to install a blocked extension and confirm an installed copy is turned off. Then apply the same settings to the larger units.

Long lists: adding hundreds of IDs by hand is slow and error prone. Google's Chrome Policy API can set extension policies per organizational unit from a script, which lets a developer apply our CSV in one run. Check Google's current API reference before you build it.

Chrome policy extension settings behind the console

The console writes Chrome policies for you. On Windows and Mac browsers that you manage with Group Policy, Intune or Jamf instead, you set the same controls in the ExtensionSettings policy. This table maps one to the other.

What you wantGoogle Admin consoleExtensionSettings value
Stop one AI extensionInstallation policy: Blockinstallation_mode set to blocked under the extension ID
Block everything not approvedBlock all apps, admin manages allowlistA default * entry with blocked, then allowed IDs listed
Install an approved tool for everyoneForce install or Force install + pinforce_installed, which users cannot turn off or remove
Install but let users turn it offCheck the current console for an equivalent optionnormal_installed
Refuse risky permissionsPermissions and URLs, blocked permissionsblocked_permissions
Keep extensions off sensitive sitesPermissions and URLs, blocked hostsruntime_blocked_hosts, with runtime_allowed_hosts as exceptions

ExtensionSettings overrides the older ExtensionInstallBlocklist and ExtensionInstallAllowlist policies. A blocked mode can carry a custom message of up to 1,000 characters, which is a good place to link your AI use policy.

Managing Windows browsers outside Google? See blocking extensions with Group Policy and the registry, extensions in Intune and the Edge extension blocklist. The same ID list works in all three.

"Added by policy" and "blocked by your organization": what users see

Searches like these usually come from users who hit a policy, or admins who inherited one. Each message points to a different control.

Chrome extension added by policy

An admin force installed the extension. Chrome marks it as managed by the organisation and hides the remove and turn off controls. The user cannot change it; the admin changes the installation policy.

Disable chrome extension installed by enterprise policy

Change Force install to Allow install or Block in the console, or remove the ID from ExtensionSettings or ExtensionInstallForcelist. Then reload policies at chrome://policy on a test device.

This site is blocked by your organization's policy

For a browser agent such as Claude in Chrome, this can come from the agent's own admin site restrictions, from Chrome's URL blocking or blocked hosts, or from categories the agent refuses by design.

The maker of Claude in Chrome lets organisation admins turn the extension on or off and restrict which websites it may access with allowlists and blocklists. Our page on blocking Claude in Chrome covers that agent in detail.

Mistakes that leave AI extensions running

Most gaps we see come from small setup choices, not from missing policies. Check these four before you call the job done.

Blocking by name instead of ID

Store names change and copies reuse them. The ID is the only stable handle, so every entry in your console should be an ID you have checked.

Forgetting a group override

A group with its own app settings wins over the organizational unit. One forgotten pilot group can keep a whole department on the old rules.

Only covering Chromebooks

Chrome on Windows and Mac needs enrolment in Chrome Browser Cloud Management, or a separate policy from Group Policy, Intune or Jamf. Unenrolled browsers ignore the console.

Never revisiting the list

A block list written once is out of date within weeks. Plan a weekly review, or import a list that is refreshed for you every day.

How the AI Agent Blocklist plugs into Google Admin

We maintain the IDs so you do not have to search the store every week. You choose how strict to be; the file does the typing.

What the Chrome file contains

Chrome extension IDs of AI agent extensions, AI sidebars with page access, browser automation add-ons and companion extensions of AI browsers. Each row carries the tool name, category and risk grade.

Pick your threshold

Block every agent extension, or only the higher risk grades while you review the rest. The grade is in its own column, so a filter in a spreadsheet is enough.

Stay current

The lists update every day as new agents appear. Compare the new file with your console list once a week and add the new IDs.

fcoeoabgfenejglbffodgkkbkcdhcgfnClaude (Anthropic), the Claude in Chrome agent extension
ejcfepkfckglbgocfkanmcdngdijcgldChatGPT search (OpenAI)
eanggfilgoajaocelnaflolkadkeghjpHARPA AI, web automation with AI models
mokiolnliafiddobcgmmnmhbmimpcconSider Operator

Four example rows. The full file holds every agent extension we track, across Chrome, Edge and Firefox, with the columns you need to filter.

Block

  • Agent extensions that click, type or submit
  • Companion extensions of AI browsers
  • Sidebars that read all sites and send page text out
  • Lookalikes of approved tools

Monitor

  • Writing helpers limited to a few sites
  • Extension requests that mention AI
  • New installs in the extensions report
  • Agent domains in DNS logs

Allow

  • AI tools your company licensed and reviewed
  • Force installed tools with a data agreement
  • Approved tools pinned for the right units
  • Exceptions with an owner and end date
Extensions are one layer. Pair the console list with network blocking of agent domains, covered in Chrome AI policies in Google Admin, and turn off built-in browser AI as shown in disable Gemini in Chrome. Schools should also read AI browsers on Chromebooks.

One list for every Chrome unit you manage

The AI Agent Blocklist covers AI browsers, agents, agent extensions, coding agents and MCP servers. For Google Admin you use the extension file; the domain files go to your firewall or DNS filter.

From $99 a month for one organisation. After an order we set up your download access within one business day.

See plans Ask a question
  • Chrome and Edge extension block lists ready for the Google Admin console, Group Policy and Intune
  • CSV and JSON with tool name, category, identifier type and risk grade
  • 2,600+ AI agent domains as a plain list, hosts file and firewall list
  • 5,300+ agent identifiers on devices, including extension IDs and app IDs
  • Daily updates as new agents and extensions appear

Ask us or order a plan

Tell us what you need to block and where your rules live. We reply within one business day.

  • Files for firewalls, DNS filters, secure web gateways, MDM and browser policies
  • Updated every day as new AI browsers, agents and extensions appear
  • Plans for one organisation, for MSPs and for product makers, or a hosted filter we run for you

We use your details only to answer you. No newsletters.

Questions admins ask

Where do I block Chrome extensions in the Google Admin console?
Go to Devices > Chrome > Apps & extensions > Users & browsers, pick the organizational unit, add the extension by its ID and set the installation policy to Block. To block everything you have not approved, change the Chrome Web Store setting to Block all apps, admin manages allowlist.
What does chrome extension added by policy mean?
It means an administrator installed the extension through a Chrome policy, usually a force install from the Google Admin console or the ExtensionSettings policy. Chrome treats it as managed, so the user cannot remove or turn it off. Open chrome://policy to see which policy installed it.
How do I disable a Chrome extension installed by enterprise policy?
Only an administrator can. Change the extension's installation policy in the Google Admin console from Force install to Allow install or Block, or remove it from the ExtensionSettings or ExtensionInstallForcelist policy on Windows and Mac. A user on a managed browser cannot override it.
Why does the Claude Chrome extension say this site is blocked by your organization's policy?
Usually one of three things. Your Claude organization admin restricted which websites Claude may access, a Chrome policy such as URL blocking or runtime blocked hosts stops extensions on that site, or the extension itself refuses a high-risk category of site. Ask your IT team which control applies.
Does blocking an extension remove it from devices that already have it?
Setting an extension to Block in the Google Admin console disables it on managed browsers in that organizational unit, including copies that were installed before the policy arrived. Unmanaged browser profiles and personal Chrome installs are not covered, which is why network blocking of agent domains is the second layer.
Can I block AI extensions only on Chromebooks used by students?
Yes. Extension policies apply per organizational unit or group. Put student accounts in their own unit and use the allowlist mode there, while staff units keep a blocklist of agent extensions. Test changes on a small unit first.

Want the wider picture first? Start with how to block AI browser extensions and how to judge whether an AI extension is safe.

Google, Chrome, Chromebook, ChromeOS, Google Admin and Gemini are trademarks of Google LLC. Claude and Claude in Chrome are trademarks of Anthropic, PBC. ChatGPT is a trademark of OpenAI.

HARPA AI and Sider are trademarks of their respective owners. Microsoft Edge and Intune are trademarks of Microsoft Corporation.

Jamf is a trademark of Jamf. Gartner is a trademark of Gartner, Inc.

Apple, Mac and macOS are trademarks of Apple Inc. Windows and Microsoft Copilot are trademarks of Microsoft Corporation.

Names are used only to identify the products discussed. AI Agent Blocklist is not affiliated with, sponsored or endorsed by these companies.