How to Block ChatGPT at the Dealership, and Why Blocking Alone Does Not Work

To block ChatGPT at a dealership, block its web and app domains at the firewall or DNS filter, push the same block to store computers through device management, and pair it with a written policy and one approved AI tool.

The block is the easy part. Making it hold is the hard part, because every person in the building carries a phone.

Last updated: November 2, 2026

Back when Facebook was new, we blocked it on dealership computers. Everybody just sat there on their phones in the dealership and used it anyway. Nothing changed except where they did it. Same habit. Different screen.

I told that story at the end of our DealersEdge webinar on AI governance this October, when an attendee asked whether there are ways to block personal AI apps from employees. There are. Here’s what I said, starting at 52:28 in the recording: “There absolutely are ways to block and control that. But even without doing that, you still need to have policies in place. I think that’s key.” Then, a few seconds later: “So, blocking is, it does stuff, but it’s hard to really control that way.”

That’s still my answer. This post is the longer version: how to block ChatGPT and the other public AI tools on dealership computers and networks, what each method actually stops, what it misses, and the policy and the approved tool you need next to it so the block means something. It’s one piece of the bigger picture we cover in our guide to AI governance for dealerships.

Five Ways to Block ChatGPT at the Dealership

Blocking ChatGPT at a dealership means stopping store-owned devices and store networks from reaching chatgpt.com and the ChatGPT apps, usually with a web filter category, a DNS block, device management rules, and, on the company’s own AI plan, an allowed-workspace setting. It covers the store’s gear. It does not cover anyone’s personal phone.

OpenAI publishes its own instructions for this, which tells you how common the request is. Very common. Its corporate network controls guide says an organization that wants to stop people reaching ChatGPT, personal accounts included, through the iOS, macOS, and Android apps can block a listed set of domains at its firewall, proxy, or secure web gateway. It also describes blocking the address ChatGPT uses for logged-out, anonymous chats, and, for stores on ChatGPT Enterprise, adding a header at the network layer that only lets your company workspace load. That last one needs managed devices with the right certificates installed. It isn’t a switch your general manager flips.

Here’s how the five methods compare. Same tools, very different reach.

Look at the right-hand column. Every row has the same hole in it. The phone in somebody’s pocket, signed in with their own email, walks right around all five. Phones win.

That doesn’t make blocking useless. On store computers it does real work: it makes it much harder for someone to drag a spreadsheet of customer data into a public chatbot from the F&I office, which is the upload you’re actually worried about. A block is a speed bump on the store’s own equipment.

Why Blocking Alone Doesn’t Hold

The people using personal AI at work aren’t rebels. They’re your best salespeople and your busiest office staff, and they found something that saves them an hour. Microsoft and LinkedIn’s 2024 Work Trend Index, a survey of 31,000 people across 31 countries, found that 78% of AI users bring their own AI tools to work. Not the company’s. Their own.

Block that on the desk computer and the need is still there. It moves. The service advisor who used ChatGPT to clean up a customer email will type the same thing into the app on their phone, on cellular data, where you can’t see it and can’t stop it. Now the customer’s name and repair history went into a personal account anyway, and you’ve lost the one place you might have noticed. That’s a real problem.

Samsung learned this in public. In May 2023, Bloomberg reported that the company had banned generative AI tools for staff at one of its biggest divisions after engineers uploaded internal source code to ChatGPT. Look at what the same memo promised next, though: Samsung said it would build its own AI tools for things like translation and summarizing documents. Even a company that size didn’t think a ban could stand on its own. It needed somewhere to send people. That’s telling.

To be clear, I’m not telling you to skip the block. I use AI every day and I want dealers using it. The block just can’t be the plan. It’s the part of the plan that covers your own equipment, and it only works when the rest is in place.

Blocking without a policy is a suggestion.

Write the Rule Before You Write the Block List

The policy is what covers the phone. A filter can’t reach a device you don’t own, but a signed rule reaches the person holding it. That’s why I put it first.

In the webinar I gave the version I’d start with, at 28:04: “ban the use of any personal tools. You cannot install Claude. You cannot install ChatGPT. You may not use it in a webpage. All of that stuff is banned by personal use unless it’s installed and you’re using it on a corporate account.” When an attendee asked whether staff should sign off on it, the answer was yes, in the employee handbook, and the number one line to start with is no personal use of AI in the dealership and no dealership data in any AI the store doesn’t control.

Two sentences. That’s it. Everybody can remember them. A salesperson who pastes a credit application into a personal chatbot on their own phone is now breaking a written rule they signed, not exploiting a gap in your firewall. That changes the conversation in the manager’s office, and it changes how often it happens.

Who owns that rule? Somebody at the C level. Our guide to setting up an AI governance committee covers who sits on it and how it approves tools, and our dealership AI use policy template gives you the wording to start from. Dealers don’t like policy work. Do it anyway.

The rule reaches places the firewall never will.

Give People Somewhere to Go

A ban with no alternative fails the same way the Facebook block did. In the webinar, right after the ban line, I said it straight: “You can’t just say nobody’s allowed to use AI. You want them to use AI, but then you need to restrict how it’s used.” If the only AI that works at your store is the one on a personal phone, that’s the one people will use.

Brandon Dorsey, our Vice President of Technology Strategy, makes the same point from the other direction. His view is that the effort shouldn’t go into blocking what staff will use anyway. It should go into deciding the store’s AI strategy and its approved platforms, and then giving people a sanctioned path, because without one they go off to personal tools.

A sanctioned path means one business AI plan the dealership owns. People sign in with their Microsoft 365 account, so multi-factor authentication carries over and a departing employee loses access the day their account is turned off. Training on your data is off for the whole organization. Somebody can see who’s using it and for what. Any of the big three works. Which product you pick matters less than those controls, which is the whole argument of our Copilot vs ChatGPT vs Claude comparison for dealerships, and turning off training is its own topic in does ChatGPT train on your data.

Some dealers would rather not run that themselves. Helion’s Secure AI Chat for dealerships is a private, managed AI workspace with the leading models in one place, role-based access, adoption reporting, and the same workspace on staff phones, set up and run as part of Helion’s managed AI services. That phone piece matters here. It puts a governed option on the device the block can’t reach.

Once people have a good tool, the block stops feeling like a punishment. It’s just how the store works.

What About the Personal Accounts Already in Use?

This was the question I didn’t have an answer for. Near the end of the webinar at 50:05, an attendee asked how to wipe out all the personal accounts, and I gave the honest answer: “I’d have to get back to you on that because I don’t actually have the answer to that.” I said Brandon had thoughts on it.

He did, and he wrote them up in the committee guide. You can’t reach into an employee’s personal ChatGPT or Gemini account and pull the dealership’s data back out. What you can do is make the personal account unnecessary. It’s blunt. It’s right. That matches what I said on the same call about data that already went in: “And then at that point, it’s gone.”

There’s one more lever, and it’s narrow. Some business plans can take over accounts opened with your company’s email domain. Anthropic’s help center says Claude Enterprise admins can find, claim, and migrate existing personal Claude accounts on a verified company domain, and that a Team plan can block new personal accounts on that domain without claiming old ones. OpenAI’s domain verification works with ChatGPT Enterprise so people with a matching work email can be moved into the company workspace. Neither one touches an account somebody opened with their personal Gmail address. Anthropic’s own help page says so.

So use the domain claim if your plan has it. Then assume there are other accounts you’ll never see, and let the policy and the approved tool do the rest.

The old accounts don’t get deleted. They get abandoned.

The Order We’d Do It In

We’ve already seen this exact problem at stores of every size, across the 2,000-plus dealerships and 35,000 end users we support. The order matters more than the tools. Here’s the sequence we’d run at your store.

1. Find out what’s in use. Ask managers, and pull the firewall’s web logs for the AI domains. The numbers usually surprise people, and they tell you which departments need the approved tool first.
2. Write the rule and get it signed. No personal AI for dealership work, no dealership data in any AI the store doesn’t control, in the employee handbook with a signature.
3. Turn on the approved tool. One business plan, sign-in through Microsoft 365, training off, licenses handed out by group.
4. Block the rest on store equipment. The generative AI category at the firewall, a DNS block for the main AI domains, and device management rules against AI app installs and unapproved browser extensions.
5. Claim the work-email accounts. If your plan supports a domain claim, use it, and tell people before it happens.
6. Train people, then check the logs again. A short session on the rule and the approved tool, and a regular look at what the filter is still catching.

Notice where the block sits. Fourth. Before it, people have a rule and a tool. After it, you’re checking whether the plan is working.

What Dealers Ask About Blocking AI

Block the Store’s Gear, Govern the People

The firewall rule is the quick part. The policy takes a meeting and a signature. The approved tool takes a decision somebody at the top has to make. Dealers who do only the first one end up exactly where we ended up with Facebook: a block on the desk, the same behavior on the phone, and a little less visibility than before. Don’t stop there.

If you’re not sure what’s running at your store today, start with a complimentary IT and cybersecurity assessment. We’ll look at which AI tools your people are using, what your network and devices can block now, and what it would take to put everyone on one governed tool.

A block covers your equipment. A policy and a good tool cover your people.