Block AI on FortiGate: AI sites, AI browsers and agents
Staff are installing AI browsers and agent extensions that act with their logins, and new ones appear every week. A FortiGate can stop them at the edge if it has a list to work from. Here is how to give it one.
Security Fabric > External Connectors > Threat Feeds
Next step
From $99 a month for one organisation
AI Agent Blocklist for FortiGate
2,600+ AI agent domains as a threat feed file your FortiGate pulls on its own schedule, updated every day.
What FortiGate admins are being asked to block
The requests used to be "block ChatGPT". Now they are about AI that acts on its own. Gartner advised security leaders in December 2025 to block AI browsers for now, and the tools keep multiplying.
AI browsers
Comet, ChatGPT Atlas, Dia, Opera Neon and agent browsers that browse, click and buy for the user.
Autonomous agents
Web agents such as manus.im and genspark.ai that take a task and work through sites alone.
Agent extensions
Extensions that read every page and act in the user's open sessions, including mail and banking tabs.
Meeting agents
Note-takers such as Read AI at read.ai that join calls and send transcripts out.
Three ways to block AI on a FortiGate
FortiOS gives you three places to put a list of AI domains. Most teams use the first two together: the web filter for traffic through policies, the DNS filter for every lookup.
| Method | Where you set it | Updates | Inspection needed | Use it for |
|---|---|---|---|---|
| FortiGuard Category threat feedMain method | Security Fabric > External Connectors, then a Web Filter profile | Pulled from a URL on a schedule | Certificate inspection to match host names; deep inspection for full URLs | The full AI agent list in web filtering |
| Domain Name threat feed | Security Fabric > External Connectors, then a DNS Filter profile | Pulled from a URL on a schedule | None, works on DNS queries | Blocking lookups for every device, including guests |
| Static URL filter | Web Filter profile > Static URL Filter | Typed by hand | Deep inspection for paths | A few one-off entries, such as Google AI Mode |
Step by step: add the AI list as a FortiGuard Category threat feed
These steps follow the FortiOS 7.4 and 7.6 administration guides. Labels are close in other versions; if one differs, check the current admin console.
Create the feed
Go to Security Fabric > External Connectors, click Create New and choose FortiGuard Category under Threat Feeds.
Point it at the list
Enter a name and paste the download URL we give you into URL of external resource. Turn on HTTP basic authentication with your credentials.
Set the refresh
Refresh Rate is in minutes, from 1 to 43,200, default 5. Our list changes daily, so 60 is plenty.
Check the entries
Save, edit the feed again and click View Entries. The Last Update field shows when it was pulled.
Prefer the CLI? The same feed looks like this. The category number must be free; category and domain feeds share the range 192 to 221.
Block the feed in a Web Filter profile
The feed does nothing until a profile uses it and a firewall policy uses the profile. Do both, and turn logging on so you can see the first blocks.
Web Filter profile
- Open Security Profiles > Web Filter and edit the profile your user policies use.
- Enable FortiGuard Category Based Filter.
- Find the Remote Categories group. Your feed is listed there by name.
- Set it to Monitor for the first week, then to Block.
- Click OK.
Firewall policy
- Open Policy & Objects > Firewall Policy and edit the outbound policy for users.
- Under Security Profiles, enable Web Filter and pick the profile.
- Select an SSL inspection profile. Certificate inspection is enough to match host names.
- Enable Log Allowed Traffic so monitor hits appear in the logs.
- Save and test from a client on that policy.
Block the same list at DNS with a Domain Name feed
DNS filtering catches lookups before any connection opens, and it works for devices that never pass through a web filter policy. Use the DNS filter import file from our download here.
Create the domain feed
- Security Fabric > External Connectors > Create New, then Domain Name.
- Paste the download URL of our DNS filter file, add the credentials and set the Refresh Rate.
- The file has one domain per line. Simple wildcards are supported.
Use it in a DNS Filter profile
- Security Profiles > DNS Filter, edit the profile.
- Enable FortiGuard category based filter; the feed appears under Remote Categories.
- Choose Monitor first, then Block. Apply the profile in your firewall policy.
FortiGate limits: plenty of room for an AI list
Threat feed limits are set per device, across all feeds, and depend on the model class. An AI agent list of a few thousand entries uses a small share of them.
Entries above the limit are not loaded and the FortiGate shows a warning. File size limits run from 32 MB on entry-level models to 128 MB on high-end ones.
Block Google AI Mode and Gemini on FortiGate
"Block google ai fortigate" usually means one of three things. Each needs a different rule, because Google runs several AI features on hosts you cannot block outright.
The Gemini app
The Gemini web app has its own host, gemini.google.com, so a URL or DNS rule can block it without touching Search.
Clean block by hostGoogle AI Mode
AI Mode runs on google.com itself. With deep inspection, a Static URL Filter entry of type Wildcard that matches the AI Mode search parameter blocks it. Fortinet's knowledge base describes this pattern.
Needs deep inspectionAI Mode in Chrome
For managed Chrome, the AIModeSettings policy set to Disabled removes the AI Mode entry points in the address bar and New Tab page.
Pair it with the URL ruleHow to block Chrome extensions with FortiGate
A firewall sees connections, not the browser's extension list. So FortiGate cannot remove an extension, but it can make an AI extension useless and keep new installs down.
Block what the extension calls
Agent extensions need their backend. Our list includes those domains, so a blocked backend means the extension stops working.
Works with the feed aboveBlock the Chrome Web Store
Application Control has a signature for Chrome Web Store access. It needs deep inspection and blocks every extension install, approved ones too.
Blunt, use with careBlock the extension ID
The browser policy blocks the extension itself, such as cecngibhkljoiafhjfmcgbmikfogdiko or fcoeoabgfenejglbffodgkkbkcdhcgfn.
The real fix, on the deviceWhat to block first on FortiGate
Every entry carries a category and a risk grade from Low to Severe. Most FortiGate admins start with the categories in dark blue and keep the amber ones in monitor mode for a few weeks.
- AI browsers and browser agents
- General autonomous agents
- MCP servers and integrations
- Personal AI assistants with mail access
- Autonomous coding agents on servers
- AI RPA
- Deep-research agents
- Meeting and voice agents
- AI sales agents
- Agent-building platforms
- Multi-agent orchestration
Check that it works
Five minutes of testing saves a week of tickets. Run these on a client behind the policy you changed.
Quick tests
- Open an AI browser's website from the list: the FortiGate block page should appear
- Look up the same domain with nslookup behind the DNS filter: it should be blocked or redirected
- Open an approved AI tool: it should still work
- Open google.com: Search should work normally
Where to look
- Web filter and DNS query events in the security event logs under Log & Report
- The feed's Last Update time under External Connectors
- View Entries, to confirm a domain is in the loaded list
- The policy hit count, to confirm traffic uses the policy
Troubleshooting: when the feed or the block does not work
Most problems with threat feeds come down to the download, the profile or the policy. Work through them in that order.
The feed shows no entries
Click View Entries and check Last Update. If it never updated, test that the FortiGate can resolve and reach the download host from its own management interface, and recheck the user name and password.
Certificate errors on the download
With server identity checking on, the FortiGate must trust the certificate of the download server. Check the system time and the CA store before you turn the check off.
The feed is loaded but nothing is blocked
Confirm the Remote Category is set to Block in the profile that the policy really uses, and that the policy is the one matching the client's traffic. The policy hit counter tells you quickly.
Blocks stopped after a firmware upgrade
Check the release notes for remote category matching in flow-based policies. Proxy inspection mode is the documented workaround on affected builds.
The category number is already in use
Category and domain feeds share the numbers 192 to 221. Pick a free number in the CLI, or let the GUI choose one for you.
An approved AI tool is blocked
Add the tool's domain to the Static URL Filter with the action Allow or Exempt, so it is matched before the category check. Keep each exception in a short register with an owner and an end date.
Users on laptops away from the office are not blocked
The FortiGate only filters traffic that passes through it. Use a full tunnel VPN for remote users, or apply the same list on the device with a DNS agent or browser policy.
One feed, refreshed every day
Point your FortiGate at the list once and it stays current. We set up your download access within one business day of your order. MSPs can get separate files per client.
- Firewall external list format
- DNS filter import files
- CSV and JSON with risk grade
- Risk grade on every entry
- Extension block lists
- Updated every 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
FortiGate and AI blocking: questions
How do I block AI websites on FortiGate?
Create an external threat feed of type FortiGuard Category under Security Fabric, External Connectors, pointing at the AI list. Then block it under Remote Categories in a Web Filter profile and apply the profile in your firewall policy.
What is a FortiGate block list or threat feed?
A threat feed is a text file of domains, URLs or IP addresses that the FortiGate downloads from a URL on a schedule. It appears as a remote category in web filter and DNS filter profiles, so you block it like any other category.
How do I block Google AI Mode on FortiGate?
AI Mode runs on google.com, so a domain block would take Search with it. With deep inspection, a Static URL Filter entry of type Wildcard that matches the AI Mode search parameter can block it.
For managed Chrome, the AIModeSettings policy turns off the AI Mode entry points in the address bar and New Tab page.
How do I block Chrome extensions in FortiGate?
A FortiGate cannot see installed extensions. It can block the services an AI extension calls, which stops it working, and with deep inspection it can block the Chrome Web Store.
To block the extension itself, use the browser's extension block list in Group Policy, Intune or the Google Admin console.
How often does the FortiGate refresh the AI list?
The Refresh Rate is set in minutes, from 1 to 43,200, with a default of 5. Our list changes once a day, so a refresh every 60 minutes is plenty.
How do I block Read AI or other meeting bots on FortiGate?
Meeting assistants such as Read AI are in the meeting and voice agent category of our list. Block the list in your Web Filter or DNS Filter profile, or keep that category in monitor mode first if teams still use the tool. For Teams and Zoom settings, see AI bots in Teams meetings.
Does the AI list replace FortiGuard categories?
No. The list is added as a remote category next to the FortiGuard categories you already use. Nothing in your existing category settings changes.
Fortinet, FortiGate, FortiOS and FortiGuard are trademarks of Fortinet, Inc. Google, Chrome, Gemini and Google Admin console are trademarks of Google LLC. Microsoft, Intune and Edge are trademarks of Microsoft Corporation.
Perplexity and Comet are trademarks of Perplexity AI, Inc. ChatGPT and ChatGPT Atlas are trademarks of OpenAI. Claude is a trademark of Anthropic, PBC. Dia is a trademark of The Browser Company of New York.
Opera Neon is a trademark of Opera Norway AS. Manus, Genspark and Read AI 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.