A one-page AI usage policy template for a small team
Short answer: Here is a one-page AI policy you can copy below. An AI usage policy is a short written rule set that tells staff which AI tools they may use, what information they must never paste in, who checks AI output before it is used, and who to tell when something goes wrong. For a team of 5 to 50, one page is enough. Fill in the blanks, share it, and ask everyone to confirm they have read it. It is a starting point, not legal advice.
You need a simple AI policy your team will actually read, not a 20-page document, and you want to copy it today.
I'm Ranjit, and I run Blur Studio: 15+ years building software, 120 apps taken live, and clients who have gone on to raise $10M+. I build AI systems and processes that help businesses do a lot more with a lot less time. I also work differently from most of what else is out there, more on that further down this page. Say hello.
If you want help deciding which AI jobs your team should do first, the AI Strategy Sprint finds them in 3 days.
The one-page template
Copy this into a document, fill in the brackets, and keep it to one page. Anything in brackets is yours to decide. Plain words matter more than legal tone: a policy people understand gets followed. Once the policy is written, how to train employees to use AI covers getting staff to follow it.
[Company name] AI Use Policy
Version: [date] Owner: [name and role] Review date: [6 months from now]
1. Why we have this. AI tools can save us time. They can also leak information and state wrong things confidently. This policy lets us use them safely.
2. Approved tools. You may use only the tools on our approved list, on the company account, not a personal one. Current list: [tool, what it is used for, who approved it, date]. To add a tool, ask [owner]. If staff are already using AI without asking, see what to do when staff already use ChatGPT at work.
3. Never put these into an AI tool:
- Customer or staff personal details beyond a first name: [add: health details, ID numbers, home addresses]
- Card numbers, bank details, passwords, access codes
- Anything under a confidentiality agreement or marked confidential
- Unpublished financials, contracts and legal documents: [change if your plan allows it]
- [Add anything specific to your trade]
If you are unsure, ask [owner] before you paste.
4. Check before you use. AI output is a draft. A named person reads it before it goes to a customer, is used in a contract or quote, changes a record, or is published. You are responsible for what you send, whether or not AI wrote it. Check names, numbers, dates and any claim of fact.
5. Say when AI helped. We tell [customers or clients] when [AI helped with X] if they ask, and where a contract or law requires it. We do not use AI to pretend to be a person on a call or chat.
6. Images, voices and likenesses. We do not create or use AI images, voices or videos of real people without their written consent. We follow our AI creative tool policy for [marketing and design work].
7. If something goes wrong. If you pasted something you should not have, or you sent AI output that was wrong, tell [owner] the same day. You will not be blamed for reporting quickly. We will [delete the item where we can, tell the affected person, record what happened].
8. Questions and changes. We review this policy every [6] months. Ask [owner] anything.
I have read and understood this policy. Name: ______ Date: ______
How to fill it in, section by section
Approved tools
Start with what your team already uses. Ask everyone for a day or two which AI tools they have tried, and do not punish the answers, because you want to hear the truth. Then pick a few. The most important thing to check about each is what the vendor does with what you type in. Business plans often differ from free plans. For example, Anthropic, the company that makes Claude, says on its support page that it does not use inputs or outputs from its commercial products, such as Claude for Work and its API, for model training by default. Always read the current terms of each tool on your list. My guide on keeping your data out of AI training has the questions to ask. Write the answer in your register: tool, plan, training default, who approved.
The never-paste list
This is the most useful part. Keep it short enough to remember. The US Federal Trade Commission's guide to protecting personal information tells businesses to keep only the personal data they need and to know where it lives. Your never-paste list follows the same idea: if you would not email it to a stranger, do not paste it into a tool you have not approved. If you handle health, financial or children's data, ask a lawyer what extra rules apply.
The review rule
This is where policies usually fail, because "check AI output" is vague. Tie it to real jobs. Here is a made-up register for a 12-person property firm:
| Job | AI may | Who checks | Before |
|---|---|---|---|
| Reply to enquiries | Draft | The agent who owns the lead | Sending |
| Meeting notes | Summarize | The meeting owner | Filing |
| Listing descriptions | Draft | Marketing lead | Publishing |
| Contracts and legal text | Nothing without approval | Owner | Any use |
The point: each job has a named checker and a moment when the check happens. NIST's AI Risk Management Framework, a voluntary guide from the US standards body, has four functions: Govern, Map, Measure and Manage. Your register is a very small version of Map (where we use AI) and Govern (who is accountable). The OWASP project, which publishes security guidance for AI applications, also lists misinformation and excessive agency, meaning AI given too much freedom to act without oversight, as top risks. A review step is the cheapest answer to both.
A 30-minute rollout plan
- Minutes 0 to 10: fill the brackets. Pick an owner. Keep the never-paste list to eight items or fewer.
- Minutes 10 to 15: list approved tools and check each tool's data terms.
- Minutes 15 to 25: hold a short team talk. Read the page aloud, take questions, and ask what jobs people want AI to help with.
- Minutes 25 to 30: collect sign-offs. Put the date to review in the calendar.
The estimate assumes one tidy document and a small team. Your first pass at the never-paste list may take longer. That is fine.
Keeping it alive after day one
A policy only works if it stays current. Put the review date in the owner's calendar. When someone asks to add a tool, answer within a week, or people will use it quietly. When a mistake is reported, thank the person and fix the rule that allowed it. Once a quarter, ask the team which rule gets in the way, and change it if the reason is good. A short policy that people trust beats a long one they work around.
Mistakes to avoid
- Banning AI. Staff will use it anyway on personal accounts, where you have no control. A clear approved route beats a ban.
- Copying a policy built for a big company. Dozens of clauses nobody reads are worse than eight lines everybody knows.
- No owner. A policy with no named person is a document, not a rule.
- No review date. Tools and terms change. Re-check every six months.
- No link to real work. If the policy never mentions your actual jobs, people will not connect it to what they do.
Where this fits with your other documents
A policy says what is allowed. A standard operating procedure says how to do a task step by step, and a checklist is the short version people tick off. My guide on the differences between an SOP, a work instruction, a policy and a checklist explains where each belongs. Teach the policy during onboarding with a short employee training plan. If you are planning your first AI project, my AI implementation plan shows what to do after the policy is in place.
When to get a lawyer
This page gives general information, not legal advice. Ask a lawyer before you finalize the policy if you:
- Handle health records, children's data or financial account data.
- Work in a regulated field such as law, finance, healthcare or education.
- Have contracts with clients that restrict where their information may be processed.
- Operate in the EU, UK or another place with its own AI or privacy rules.
- Plan to use AI for hiring, firing or pay decisions.
A short policy is quick for a lawyer to review, and the cost is small next to a client data problem.
How I'd approach this
I would start with eight lines and a register of approved tools, and I would write the never-paste list with the team, not for them. I would also pick one real job, such as drafting enquiry replies, and attach the review rule to it so the policy becomes a habit and not a PDF. Then I would look at the policy again in three months and fix what people ignored.
If you want to find the first jobs worth giving to AI, the AI Strategy Sprint is a paid 3-day first step where you and I find where AI could transform the business and generate new value, and you get the exact steps written down. It begins with a free 15-minute AI brainstorm call.
Questions people ask
Does a small business really need an AI policy?
If anyone on your team uses an AI tool, yes. Staff may already be using AI, and a short rule about what never goes in protects you. One page is enough for most teams under 50.
What should an AI policy include?
Approved tools, a never-paste list, a review rule, a way to report mistakes, and a named owner with a review date. The template above has all five.
Can I use this template as it is?
It is a starting point. Fill in your own tools and data rules, and ask a lawyer to review it if you work in a regulated field or handle sensitive data. It is not legal advice.
How do I get staff to follow it?
Write the never-paste list with them, tie the review rule to real jobs, and make reporting mistakes safe. Teach it in onboarding using an employee training plan.
Which AI tools are safe for business data?
It depends on the plan and the vendor's current terms, not the brand. Check whether the plan trains on your inputs by default, who can see chats and where data is stored. See my guide on AI tools that do not train on your data.
More in this topic: Problem solving: AI strategy and adoption (24 guides) →
Sources
- NIST: AI Risk Management Framework
- NIST: AI RMF Playbook
- NIST: AI RMF Core (Govern, Map, Measure, Manage)
- OWASP Gen AI Security Project: Top 10 for LLM Applications
- FTC: Protecting Personal Information, A Guide for Business
- FTC: Data security guidance for business
- Anthropic: Is my data used for model training?