Your staff are almost certainly using AI already, and some of them are doing it from a personal account. An AI policy template gives you something to hand them: one short page that says which tools are approved, what must never be pasted in and who checks the work.
Below is an AI acceptable use policy for a small business that you can copy, change and issue this week. It is a starting point, not legal advice. Read it with whoever looks after your contracts and data protection before it goes to staff.
How to use this template
Copy the eight sections into a document. Replace anything in [square brackets]. Cut anything that does not apply, and keep the finished policy to two pages.
Two things make it work. First, the policy must name real tools. A rule that says “use approved tools” with nothing approved sends people straight back to personal accounts. Second, one named person owns it. If nobody owns it, nobody updates it.
The AI usage policy template
[Company name] AI usage policy
Version 1.0, [date]. Owner: [name and job title].
1. Purpose and who it covers
This policy covers everyone who does work for [company name], including staff, contractors and freelancers. It applies whenever you use an AI tool for company work. An AI tool is any chat assistant or app that writes text, images, code or summaries from a prompt.
2. Approved tools
Use only these tools for company work: [tool and plan, for example “Claude Team”], [tool and plan]. Sign in with your company account, never a personal one. To ask for another tool, contact [name]. Do not add AI browser extensions, meeting recorders or note-taking apps without approval.
3. What must never go in
Never paste or upload:
- passwords, security codes, API keys or payment details
- customer or staff personal data, such as names with contact details, health details or HR records, unless [name] has said the approved tool may be used for that task
- contracts, prices or plans marked confidential, or anything a client has asked us to keep private
- unpublished accounts or financial results
If you are not sure, leave it out and ask [name]. Remove names and identifying details wherever you can.
4. Check everything before it leaves the building
AI gets things wrong and sounds certain when it does. A person checks facts, figures, quotes, prices and any legal, tax or medical statement before it goes to a customer or is published. You are responsible for what you send, whoever or whatever drafted it. A person, not an AI tool, makes decisions about hiring, pay or discipline.
5. Customer and personal data
Use personal data in an AI tool only for the purpose we collected it for, and only in an approved tool cleared for it. [Name] checks that our privacy notice still matches how we use AI. Do not use AI to build profiles of customers or staff without asking [name] first.
6. Accounts and passwords
Use your own login and never share it. Use [single sign-on / a unique password and two-step verification]. Do not connect an AI tool to your email, files or accounting software without approval from [name]. [Name] removes a leaver’s access on their last day.
7. If something goes wrong
If you paste something you should not have, or AI-written work with a mistake reaches a customer, tell [name] the same day. Do not delete anything or try to fix it alone. We will decide whether a personal data breach has happened and whether we must report it. We would much rather hear about a mistake early than find out late.
8. Owner and review
[Name] owns this policy and answers questions about it. We review it every [six] months and whenever we add a tool. Breaking this policy may be dealt with under our [disciplinary procedure].
How to roll it out
A policy on its own changes little. This is the order that works for a small team.
- Set up the tools first. Create company accounts for the one or two tools you are approving, switch on single sign-on if the plan offers it, and check each plan’s data settings. The policy should describe your settings, not the other way round.
- Fill in the brackets. Put real names and real tool names in. A policy with brackets left in looks abandoned.
- Run a 30-minute session. Walk through section 3 with examples from your own work: a customer email, a supplier quote, a staff rota. Show what can go in and what cannot.
- Get a yes from everyone. Ask each person to confirm by email that they have read it, and keep the replies.
- Make asking easy. Name who approves a new tool and promise an answer within a couple of working days. If the answer takes a month, people stop asking.
- Review it after a month. Ask what people wanted to do and could not. Change the policy, add a version number and tell the team.
UK data protection: what to read next
AI tools are covered by the same UK data protection law as any other way of handling personal data. This template does not explain that law and we are not going to paraphrase it. The ICO, the UK regulator, does that better.
- ICO: artificial intelligence is the starting point. Its guidance on AI and data protection is longer, and the ICO notes that it is under review because of the Data (Use and Access) Act, so check you are reading the current version.
- ICO advice for small and medium organisations has plain-English tools for smaller firms.
- The ICO’s guide to personal data breaches says certain breaches must be reported to the ICO within 72 hours of becoming aware of them, where feasible. That is why section 7 asks for the same day.
If you handle health, financial or other sensitive data, or work in a regulated sector, take advice from a solicitor or your regulator before you rely on any template, including this one.
Want the tools set up to match the policy?
A policy only works if the tools behave the way it says. Secure AI Onboarding is our service for that. We set up your firm’s AI tools (Claude, ChatGPT, Microsoft Copilot, Google Gemini or Grok), lock them down properly, write the usage rules with you and train your staff.
It starts from £495 plus VAT for up to five people, with one tool set up securely, written usage rules and a remote training session. The tool subscriptions are paid to the vendor and are not included.
See AI training for business or tell us what you need. If you are still choosing a tool, start with which AI assistant your business should use or the Claude for small business guide.
What people ask about this.
Does a small business need an AI policy?
UK data protection law applies to personal data in AI tools whether or not you have a written policy, and your staff will use these tools either way. A short policy is the quickest way to get everyone following the same rules. It does not replace advice on your own situation.
How long should an AI acceptable use policy be?
One to two pages. Staff read short rules and skip long ones. Put tool detail, such as which settings to use, in a separate one-page guide that you can update without reissuing the policy.
Can staff use their personal ChatGPT or Claude accounts for work?
We would say no. A personal account has data settings you cannot see or change, and you cannot remove a leaver's access. Business plans give an owner central controls, and Anthropic and OpenAI both say on their own pages that they do not train on business-plan data by default. Check the plan you buy.
How often should we review the policy?
Every six months, and whenever you add a tool or change a plan. AI products change quickly, and a policy that names last year's settings is one people stop trusting.
Your staff are probably using AI already, on personal accounts, with no rules. We set your tools up properly, lock them down to the standard a large company would insist on, write down how your firm uses them and train your people on their own work. Chris Leah, who runs Happy Webs, has set up AI tools and trained staff in small and micro businesses in the UK. This plan is for firms of up to 100 people that are behind on AI and want to catch up, at a price that suits a firm your size. We call it Secure AI Onboarding.
Explore the serviceTell us what you want to improve →
Also relevant: Claude setup + trainingChatGPT setup + training
