How to Write a Simple AI Use Policy for Your Business
By Adam, AI Engineer · 3 November 2026 · 5 min read
Short answer: an AI use policy is a one-page document that says which AI tools your staff may use, what information they must never put into them, who checks the results, and who to tell when something goes wrong. A small business can write a useful one in an afternoon. You do not need a lawyer to start, though you should get legal advice if you handle health, financial or children's data.
Why you need one now
Most staff already use AI tools, often free ones on their own accounts. Without rules, private customer data can end up in tools you have never checked, and nobody knows who is responsible for mistakes.
Big companies are tightening up. In November, Goldman Sachs limited AI coding tools such as Claude Code to its engineers, citing productivity and risk management, according to Bloomberg. And the Northflank breach showed how a stolen AI key can be used to run up someone else's bill. A short policy helps you avoid both kinds of trouble.
The six parts of a good AI policy
1. Approved tools. List the tools staff may use for work, and say which account to use. Business plans often give you more control over data than free personal accounts, so check each tool's terms. Anything not on the list needs approval first.
2. Who can use what. Not every tool suits every person. Writing assistants are low risk for most staff. Tools that can run code, send emails or change systems on their own (AI agents) should be limited to people trained to use them.
3. Data rules. This is the most important part. Use three simple levels:
| Data type | Examples | Rule |
|---|---|---|
| Public | Website text, published prices | Fine in any approved tool |
| Internal | Plans, internal emails, draft proposals | Approved business accounts only |
| Private | Customer details, passwords, ID numbers, health or payment data | Never, unless a named tool is approved for it |
4. Checking the output. AI tools can be confidently wrong. Say that a person must check any AI output before it goes to a customer, is published, or is used for a decision about money, hiring or the law. The person who sends it is responsible for it.
5. Keys, accounts and spending. If your business uses AI through an API, keep keys in one place, give each project its own key, set spending limits where the provider allows, and know how to cancel a key quickly. Remove access the same day someone leaves.
6. Reporting problems. Tell staff who to contact if they paste something they should not have, see a strange bill, or get a harmful result. Make it clear that reporting quickly will not get them in trouble. Problems you hear about early are cheaper to fix.
A simple template
You can copy this and fill in the blanks:
[Company] AI use policy
- Approved tools: [list], using company accounts only.
- AI agents and coding tools: only [roles], after training.
- Never enter: customer personal data, passwords, payment details, [other].
- A person checks all AI output before it leaves the company.
- API keys are managed by [name], with spending limits.
- Report problems to [name or email] straight away.
Reviewed every [6] months by [name].
Mistakes to avoid
- Banning AI completely. Staff often keep using it on their phones, and you lose all visibility.
- Writing ten pages. Nobody reads them. One page that people follow is better.
- Never updating it. Tools and prices change fast. Put a review date in the policy.
Bottom line
Start small: list your approved tools, ban private data in unapproved ones, and name one person in charge. You can add detail later. A simple policy that everyone knows is far more useful than a perfect one that sits in a folder.