Block AI sites on Sophos Firewall, WatchGuard and UniFi

AI browsers and agents are showing up on small and mid-size networks too, and they act with your users' logins. These three platforms can all block them, but each one takes a list differently. Here is where an AI agent list fits on each.

Next step

AI Agent Blocklist for Sophos, WatchGuard and UniFi networks

2,600+ AI agent domains as a plain list, CSV and DNS filter files, ready for feeds, imports and DNS filters, updated every day.

From $99 a month for one organisation

What you are blocking, and why a list beats typing

The AI tools that need blocking now are the ones that act on their own. Gartner advised security leaders in December 2025 to block AI browsers for now (what that advice means). New agents appear every week, so a list you typed last quarter is already out of date.

AI browsers

Comet, ChatGPT Atlas, Dia and others that browse logged-in sites, fill forms and buy for the user.

Cloud agents

Services such as manus.im and browser-use.com that work through websites alone.

Agent extensions

Extensions that read every page and act in open tabs. Their backends are on the list.

MCP servers

Hosted servers and directories that hand agents new tools, such as smithery.ai. See blocking MCP servers.

Your firewall's built-in web categories stay as they are. We do not describe or rate them. Our list is added next to them as a custom category, feed or blocklist, focused on agents.

At a glance: how each platform takes a list

The right method depends on whether the platform can fetch a list by itself, how many entries it holds and whether it filters by DNS or by web proxy. Check the current admin console for your version.

PlatformMethodPulls from a URLSizeSees
Sophos Firewall 21 and laterThird-party threat feed, Domain typeYes, with basic authDepends on the applianceHost names
Sophos FirewallCustom web category, External URL databaseYes, HTTP or FTP, every 48 hoursNo set limit; cache depends on RAMHost names, paths with decryption
Sophos FirewallCustom web category, Local uploadNo, file upload2,000 entriesHost names, paths with decryption
WatchGuard FireboxWebBlocker exceptionsNo, import in Policy ManagerNo documented limitDomains; full URLs with content inspection
WatchGuard, recent FirewareCustom dynamic alias with a domain feedYes, HTTP or HTTPSFile up to 2 MBDomains, used in policies and Blocked Sites
WatchGuard CloudDNSWatch domains listNo, CSV or TXT upload1,000 domains per configurationDNS names
UniFiContent Filter blocklistNo, typed or pastedNo documented limitDNS names
UniFiZone-Based Firewall, domain destinationNoNo documented limitDomains

Sophos Firewall: block AI and generative AI agents

SFOS 20, 21 and 22

Sophos Firewall gives you three places for a list. For an AI agent list of a few thousand domains, the threat feed and the custom category are the two that scale.

Third-party threat feed

Version 21 and later. Pulls a list of domains from a URL on a schedule, with Block or Monitor as the action.

Fits a daily list

Custom web category

Any current version. Upload a file of up to 2,000 entries, or point the category at an external URL database.

Works well in web policies

URL group

Domains added one at a time. Fine for approved tools or a short list, not for thousands of entries.

Use for exceptions

Option A: third-party threat feed

  1. Go to Active threat response > Third-party threat feeds and click Add.
  2. Set the Indicator type to Domain and the Action to Monitor for the first week, then Block.
  3. Paste the download URL as the external URL. Under Authorization, choose Basic authentication with your credentials, and keep Validate server certificate on.
  4. Choose a polling interval, such as 24 hours. Click Test connection, then Save.

Option B: custom web category

  1. Go to Web > Categories and click Add.
  2. Name it, choose a classification and pick the Configuration type: Local for a file upload, or External URL database.
  3. For Local, use Browse to upload a .txt file with one domain per line. Only the first 2,000 entries are imported.
  4. For External URL database, enter an HTTP or FTP address. The firewall checks it every 48 hours.
Threat feeds take plain entries only: no wildcards, ranges or regular expressions. They need the Xstream Protection licence, a firewall rule from internal to WAN, and application classification or an IPS policy on that rule. An external URL database cannot use authentication, so host the file on an internal web server if you choose it.

Sophos web policy: put the block where traffic passes

A custom category only blocks once a web policy uses it and a firewall rule uses the web policy. Order matters: rules in a web policy are checked from the top.

Add the rule

  1. Open Web > Policies, edit your user policy and click Add rule.
  2. Under Activities, clear All web traffic and select the AI category.
  3. Set the action to Block HTTP, and for HTTPS choose Use action or Block HTTPS.
  4. Move the rule below your allow rule for approved AI tools, switch Status on and save.

Attach it to the firewall rule

  1. Open Rules and policies > Firewall rules and edit the rule for user traffic.
  2. In the Web filtering section, choose the web policy.
  3. Turn on scanning of HTTP and decrypted HTTPS if you inspect HTTPS.
  4. Save and test from a client behind the rule.
Custom category entries cover subdomains automatically and can include a path, such as a single page on a site, which matches only with HTTPS decryption. For approved tools, a URL group with an allow rule above the block keeps them working.

How to block Chrome extensions in Sophos XG Firewall

This question comes up often. We found no documented Sophos Firewall setting that blocks single Chrome extensions, and a network device cannot see which extensions a browser has installed. There are still three things that work.

Block the backend

AI extensions call their own services. With those domains blocked by the feed or category, the extension stops working.

Block the extension ID

Chrome and Edge policies block extensions by ID, such as cecngibhkljoiafhjfmcgbmikfogdiko for a browser agent extension.

Allow only approved ones

Block all extensions by default and allow a short list. It is the strongest setting, with more upkeep.

Our download includes Chrome and Edge extension block lists for Group Policy, Intune and the Google Admin console. Step by step: blocking Chrome extensions with Group Policy and building a Chrome extension allowlist.

WatchGuard: block AI sites on a Firebox

Fireware 12.x, newer releases and WatchGuard Cloud

On a Firebox, WebBlocker exceptions are the classic route and work on most installs. Newer Fireware releases add domain feeds that pull a list by themselves. Cloud-managed sites can also use DNSWatch.

WebBlocker exceptions

Deny rules matched by pattern, exact match or regular expression, imported from a text file in Policy Manager.

Works on Fireware 12.x

Domain feed alias

A custom dynamic alias with an IP address or domain feed, refreshed from a URL. It needs a recent Fireware release on supported models.

Self-updating

Blocked Sites with FQDN

Good for a handful of names. WatchGuard recommends about 2,048 domain names in total.

Too small for the full list

WebBlocker exceptions from a file

  1. In Policy Manager, open Subscription Services > WebBlocker > Configure and edit the WebBlocker action used by your proxies.
  2. On the Exceptions tab, choose Import. The Web UI cannot import.
  3. Use a UTF-8 text file with one rule per line. A line can be just the pattern, which uses pattern match by default.
  4. Choose Append to keep your existing rules, or Replace for a clean weekly swap. Set the rules to deny.

Make sure WebBlocker sees the traffic

  1. WebBlocker must be enabled in both the HTTP proxy action and the HTTPS proxy action.
  2. Without content inspection, HTTPS is checked by domain only, which is enough for AI agent domains.
  3. With content inspection, full URLs can be matched as well.
  4. Save and push the configuration, then test a listed domain.
# Example lines in the import file (pattern match is the default; # check the pattern syntax for your Fireware version) manus.im/* *.manus.im/*
Exceptions are checked from the top and the first match wins. Put allow exceptions for approved AI tools above the deny rules. Query strings are not sent for categorisation, so exceptions cannot deny a single search parameter.

WatchGuard domain feeds and DNSWatch

If your Firebox runs a recent Fireware release, a domain feed saves the weekly import. Cloud-managed sites can add a DNS layer with DNSWatch.

Custom dynamic alias with a domain feed

  1. Open Firewall > Aliases in the Web UI, or Aliases in WatchGuard Cloud device configuration.
  2. Add a custom dynamic alias with an IP Address/Domain Feed member.
  3. Enter the Host URL, the Refresh Interval in minutes (default 60) and the authentication type.
  4. Use the alias in the To list of a deny policy, or in Blocked Sites. Each sync replaces the whole list.

DNSWatch block list

  1. In WatchGuard Cloud, open Configure > DNSWatch and your configuration.
  2. Under Domains List, add a list with List Type Block and Include Subdomains on.
  3. Upload a CSV or TXT file. Each configuration holds 1,000 domains, so split the list by category.
  4. Start with AI browsers, autonomous agents and MCP servers.
The feed file must be at most 2 MB, with one entry per line. Domains and wildcard domains are accepted; URLs with paths are not. Our plain domain list fits these rules.

UniFi: block AI sites on a UniFi gateway

UniFi Network 9.x

UniFi gateways filter by DNS and by zone policy. They do not document any import or URL loading for custom lists, which shapes the approach.

Content Filter blocklist

Under Settings, CyberSecure, Content Filter. Add domains to the Blocklist of a filter applied to your networks or clients.

Good for the top entries

Zone-Based Firewall

On Network 9.x, a policy can block a destination by domain. Useful for a few named services per zone.

Small lists only

A DNS filter on the network

Pi-hole or AdGuard Home loads the full list from a URL. Hand it out as the DNS server through DHCP.

Fits the full list

Content Filter steps

  1. Open Settings > CyberSecure > Content Filter and create or edit a filter.
  2. Apply it to the networks or clients you want to cover.
  3. Add the highest risk AI agent domains to the Blocklist, filtered from our CSV by risk grade.
  4. Save, then test from a client on that network.

Full list through a DNS filter

  1. Run Pi-hole or AdGuard Home on a small server or container.
  2. Add our DNS filter file as a list by URL so it refreshes by itself.
  3. In the UniFi network settings, hand out the filter's address as the DNS server.
  4. Block outbound DNS to other resolvers with a policy, so clients cannot skip it.
Simple App Blocking under CyberSecure blocks apps and app categories chosen by UniFi; it does not take your own domain list. The full DNS route is covered in the Pi-hole and AdGuard AI blocklist guide and blocking AI with DNS.

A rollout that works on all three

The same four stages fit Sophos, WatchGuard and UniFi. Where a platform has no monitor mode, stage by network or user group instead.

Week 1

Look

Monitor action on Sophos, logging on WatchGuard, or a pilot network on UniFi. Note who uses which agent.

Week 2

High risk

Block entries graded High and Severe. Tell users what changed and where to ask for exceptions.

Week 3

AI browsers and agents

Block those categories for everyone and remove the apps on managed devices.

Week 4

Full list

Block the rest, with approved tools on an allow rule above the block.

For the general picture across firewalls, including the gaps no firewall closes, read blocking AI agents at the firewall. For AI browsers on devices, see blocking AI browsers.

One list, three platforms

Use the same AI agent list for a Sophos feed, WatchGuard exceptions and a UniFi DNS filter. The lists update every day, and we set up your download access within one business day of your order. MSPs can get separate files per client site.

SophosPlain domain list for threat feeds and categories
WatchGuardPlain list for exceptions and domain feeds
UniFiDNS filter import files, CSV for the top entries
DevicesExtension and app identifiers
UpdatesEvery day

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.

Sophos, WatchGuard and UniFi: questions

How do I block AI on a Sophos Firewall?

On SFOS 21 and later, add the AI agent list as a third-party threat feed of type Domain with the action Block, under Active threat response.

On any version, you can also create a custom web category from a file and block it in a web policy rule that is attached to your firewall rule.

How do I block generative AI on Sophos Firewall without blocking approved tools?

Block the AI agent list as a custom category or threat feed, and put approved AI tools in a URL group with an allow rule placed above the block rule in the web policy. Rules are checked from the top.

How do I block Chrome extensions in Sophos XG Firewall?

We found no documented Sophos Firewall setting that blocks individual Chrome extensions. The firewall can block the services an AI extension calls, which stops it working.

To block the extension itself, use the Chrome or Edge extension block list in Group Policy, Intune or the Google Admin console.

How do I block AI sites on a WatchGuard Firebox?

Import the AI domains as WebBlocker exceptions with the action deny, using Policy Manager, and make sure WebBlocker is on in both the HTTP and the HTTPS proxy. On recent Fireware releases you can also load the list as a domain feed in a custom dynamic alias.

How do I block AI sites on UniFi?

Add domains to the Blocklist of a Content Filter under Settings, CyberSecure, or create a zone policy with a domain destination.

UniFi documents no import, so for a full list point your clients at a DNS filter such as Pi-hole or AdGuard Home that loads the list by URL.

Does a firewall see AI traffic on HTTPS without inspection?

It sees the host name, which is enough to block AI agent domains. Full URL paths need HTTPS inspection on Sophos and WatchGuard. UniFi content filtering works at DNS level, so it sees names only.

How often should I update the AI list on these firewalls?

The list changes every day. Sophos threat feeds and WatchGuard domain feeds can poll on a schedule. For manual imports, replace the list once a week.

Sophos and Sophos Firewall are trademarks of Sophos Ltd. WatchGuard, Firebox, Fireware, WebBlocker and DNSWatch are trademarks of WatchGuard Technologies, Inc. UniFi is a trademark of Ubiquiti Inc.

Chrome and Google Admin console are trademarks of Google LLC. Microsoft, Edge and Intune are trademarks of Microsoft Corporation. Perplexity and Comet are trademarks of Perplexity AI, Inc. ChatGPT Atlas is a trademark of OpenAI. Dia is a trademark of The Browser Company of New York.

Manus, Browser Use, Smithery, Pi-hole and AdGuard are trademarks of their respective owners. Gartner is a trademark of Gartner, Inc. Names are used only to identify the products discussed. AI Agent Blocklist is not affiliated with, sponsored or endorsed by these companies.