Before You Hand AI the Keys: Building an AI Acceptable Use Policy for Your Team

Someone on your team just found an AI tool that saves them an hour a day. They signed up with a work email, uploaded a few files, and told no one. Multiply that by every employee, and you have a data problem.

In the first episode of Trust Tech Chat, Mack Slaughter and Spencer Taylor of Trust Technology Consultants took this on. Their advice fits in three words: pause and ask. This guide covers the questions to ask first and what a simple AI acceptable use policy should include.

Watch the full episode on YouTube

Why “Pause and Ask” Is the Best AI Advice Right Now

AI tools are moving faster than the rules around them. As Spencer put it, things are still a bit crazy, and not many guardrails are up yet.

The cheapest fix is a conversation before anyone signs up. Asking a few questions costs nothing. Cleaning up exposed data costs far more.

When AI Can See More Than It Should

Mack uses Microsoft Copilot as his example because it comes preloaded on so many Windows devices. Picture a business that turns it on without setting permissions.

An employee asks Copilot to pull up employment documents for the executives. If that person’s account can reach those files, Copilot will find them. It isn’t acting with bad intent. It simply has access.

AI works inside the permissions you give it. Loose permissions produce loose answers. The fix is least privilege: each person and each tool gets only the access their role needs. Review shared folders, group memberships, and file permissions before you connect any assistant to company data.

Shadow AI: Your Team Is Probably Already Using It

Shadow AI borrows from an older idea called shadow IT. Shadow IT was the personal hotspot someone plugged into the office network without asking. Shadow AI is any AI tool used for work without the company’s knowledge or approval.

Most of it starts with good intentions. An employee pastes a private spreadsheet into a free chatbot to get help fast. The data still leaves your control.

AI also shows up when nobody chose it. Google searches open with AI answers, and Spencer’s Samsung phone pushes Gemini constantly. For the bigger picture, our guide to AI and cybersecurity awareness is a good starting point.

How to Find Out What’s in Use Without a Witch Hunt

Mack has given this talk to local financial institutions and medical practices. The top question is always how to find out what’s in use. Assume AI is already there, then make it safe to say so.

  • Ask openly. A short survey about helpful tools gets more honest answers than an audit.
  • Check the network. Business-grade security platforms can inspect traffic and show which AI tools are in use. A basic consumer router can’t do this.
  • Review sign-ups and spending. Look for AI subscriptions on company cards or tied to work email addresses.
  • Offer an approved path. People drop risky tools faster when a sanctioned option exists.

Four Questions to Ask Every AI Vendor Before You Sign Up

Get clear answers to these four questions before any tool touches company data. If a vendor can’t answer plainly, that tells you something.

 

Question Why it matters What a good answer looks like
Where does our data go? Data stored outside your control is harder to protect Named storage locations and a clear security policy
Is it used to train the model? Training can fold your inputs into a shared system Training is off by default, or you can turn it off
What access does it need? Broad access raises the risk of oversharing Only the folders, apps, and accounts the task requires
What happens if we cancel? Some tools keep data long after you leave A written data retention and deletion process

 

On that last question, Mack’s advice is to read the terms of service. Confirm your data is disposed of when you exit, not just that your login stops working.

Public Tools vs. Tools You Control

Mack compares free public AI tools to standing on a public street. You have no expectation of privacy there. Treat anything you type, upload, or paste into one the same way. Spencer compares it to posting a photo on Instagram.

Tools you control change the picture. Some businesses run AI models on their own servers or in a colocation data center. A colleague of Mack’s runs assistants on three stacked Mac minis, each with its own role.

The models can reach the internet, but a firewall keeps company data from going out. Business subscriptions inside your own environment are a middle ground. When Trust merged five old customer databases into one clean list, the team ran it inside its own Copilot tenant.

What a Simple AI Acceptable Use Policy Covers

Many policy templates run dozens of pages, and nobody reads those. A practical AI acceptable use policy fits on a page or two.

Mack compares it to earlier shifts. Office computers were for Excel, not solitaire. The internet was for work, not ESPN. AI needs the same plain rules.

A simple AI acceptable use policy covers:

  • Approved AI tools. List which tools are allowed and which accounts to use, company or personal.
  • Data that stays out. Customer records, employee files, financials, patient data, and intellectual property stay out of unapproved tools.
  • Who gets access to what. Tie AI access to roles, so not everyone needs every tool.
  • Work use on work accounts. Mack’s example: don’t use the company Gemini subscription for personal images. Use your own account on your own time.
  • Human review. Someone checks AI output before it reaches a customer or a decision.
  • Spending limits. Set monthly caps on usage and credits.
  • A request process. Name who approves new tools and how employees ask for one.
  • Employee training. Walk the team through it and revisit it as tools change.

Some organizations choose no AI at all, and Mack says that’s fair too. The point is to decide on purpose.

Set Spending Limits Before the Bill Arrives

Most AI platforms now run on credits. Spencer compares it to arcade tokens: some games cost more than others.

Mack has heard stories of companies blowing through budgets because no one set a threshold. Spencer remembers the same problem with long-distance bills. Role-based access helps too. One Trust supplier partner ties AI access to each employee’s directory account.

A Safe Win: Digitizing Years of Paper Records

None of this is anti-AI. Digitizing old paper records is one of the safest wins available today.

The company Mack and Spencer bought dates back to 2002. Before an office move, helpers spent weekends scanning years of paper records and filing each by hand.

An AI model could have filed each scan automatically. Now that the records are digital, AI can search them. Need to know what circuit a customer bought in 2007? Just ask.

That lookup used to mean about twenty minutes at the file cabinet. Now it takes seconds, even when you’re working from Dallas or Lake Charles.

The biggest question in any AI rollout is still the simplest: what are you trying to accomplish? If you can’t answer that, pause and ask.

FAQ

What is an AI acceptable use policy?

An AI acceptable use policy is a short set of rules for how employees use AI at work. It covers approved tools, what data stays out, and who can access what. The goal is safe, productive use, not a ban.

What should an AI acceptable use policy include?

It should list approved AI tools, banned data types, and access by role. Add rules for human review, spending limits, and how to request new tools. Keep it short enough that employees will actually read it.

Does a small business need an AI acceptable use policy?

Yes. Small teams use free AI tools as often as large ones, often with less oversight. One upload of customer or employee data can create a privacy or compliance problem. A one-page policy sets clear expectations at almost no cost.

What is shadow AI?

Shadow AI is any AI tool employees use for work without company knowledge or approval. It usually starts with good intentions, like wanting faster help on a task. The risk is that sensitive data ends up outside your control.

Is it safe to put company information into free AI tools?

Treat free public AI tools like a public space with no expectation of privacy. Inputs may be stored, reviewed, or used for model training, depending on the provider. Keep confidential information in approved tools that run inside your own environment.

Do AI tools use your data to train their models?

Some do, depending on the provider, the plan, and your settings. Free consumer versions are more likely to use inputs for training than business plans.Always check the terms and turn training off where you can.

What questions should you ask an AI vendor before signing up?

Ask where your data goes and whether it’s used to train the model.Ask what access the tool needs and what happens to your data if you cancel.Get the answers in writing before anyone connects company data.

Talk With Trust Before You Adopt a New Tool

You don’t need to figure out AI governance alone. If your team is weighing a new tool, a quick conversation now saves cleanup later.

Trust can act as your technology consultant to vet vendors and set sensible guardrails. Our cybersecurity services help you lock down access before AI gets the keys. Reach out before your next sign-up, and we’ll help you ask the right questions.