AI acceptable use
policy template.
This template is a short staff policy for AI tools such as ChatGPT, Microsoft Copilot and Google Gemini. It lists approved tools, the data that never goes into them, how to check outputs, when to tell customers, training under the EU AI Act and what to do after a mistake.
When does a business need an AI use policy?
If your staff use a chat assistant at work, you need rules, even if you never approved the tool. The usual risk is not a dramatic failure. It is a staff member who pastes a customer complaint, a payroll file or a client contract into a personal account to get a quick summary.
A written policy does three things. It tells staff which tools and accounts they can use. It names the data that must stay out. It gives a simple route to report a mistake, so that you hear about it in time. A personal data breach can need a report to the Data Protection Commission within 72 hours under Article 33 of the GDPR, and you can only meet that limit if staff report fast.
The policy also supports your AI literacy duty. Article 4 of the EU AI Act applies to deployers of AI systems, which includes a business whose staff use AI tools for work. The 2026 Digital Omnibus on AI changed the wording to a duty to take measures that support AI literacy. Our page on AI literacy for staff explains what that means in practice.
Get the editable file.
DOCX · editable
AI acceptable use policy (Word)
We use your email to send updates to this template. We do not share it.
The policy template in full
This policy tells staff how to use AI tools at [Company name]. It applies to chat assistants such as ChatGPT, Microsoft Copilot and Google Gemini, and to AI features inside other software. Replace every field in [square brackets]. Remove the rules that do not apply to your business.
1. Scope
- This policy applies to all employees, contractors and temporary staff of [Company name].
- It applies to all AI tools that generate text, images, audio, code or summaries, on any device, when you use them for work.
- It applies to AI features inside tools that we already use, for example email, meeting software and accounting software.
- Policy owner: [Name, role]. Questions about this policy go to [email address].
2. Approved tools
Use only the tools in this table for work. Use only the company account. Do not use a personal account for work data.
| Tool | Account | Approved for | Not approved for |
|---|---|---|---|
| [Microsoft 365 Copilot] | [Company work account] | [Drafting, summarising internal documents and meetings] | [Customer health or financial data] |
| [ChatGPT, company workspace] | [Company workspace, training on our data switched off] | [Drafting, research, rewriting public text] | [Customer personal data] |
| [Google Gemini in Workspace] | [Company Google account] | [Email drafts, document summaries] | [HR records] |
| [Your tool] | [ ] | [ ] | [ ] |
To request a new tool, send its name, its purpose and its data terms to [Name]. Do not use the tool until it is in this table.
3. Data that never goes into an AI tool
Do not put these types of data into any AI tool, unless the tool is approved in section 2 for that data type:
- Personal data of customers, patients or staff (names with contact details, records, messages).
- Special category data: health, religion, ethnic origin, trade union membership, biometric data.
- PPS numbers, passport numbers and other ID numbers.
- Card numbers, bank details and account numbers.
- Passwords, API keys, access codes and security settings.
- Confidential client documents, contracts and legal advice.
- Unpublished financial results, prices or plans.
If you are not sure, remove the names and other identifiers first, or ask [Name].
4. Check every output
- You are responsible for all work that you send or publish, also when an AI tool wrote it.
- Check facts, figures, dates, names, prices and quotes against the original source.
- Do not send AI output about law, tax, health or safety to a customer before a qualified person reviews it.
- Do not use AI-generated code in live systems before a code review.
- Do not use an AI tool to make or rank decisions about people (recruitment, performance, credit) without written approval from [Name].
5. Disclosure
- If a customer deals with an AI system (for example a chat assistant on our website), tell them clearly that it is an AI system.
- Do not present AI-generated images, audio or video of real people as real.
- In [proposals / reports / client advice], tell the client where AI tools helped, if our contract or their policy requires it.
6. Records
- [Name] keeps a register of approved AI tools, their purpose, their data terms and the date of approval.
- Keep a record when an AI tool has a large part in a client deliverable or a decision about a person.
- Do not delete AI chat history that a client, a regulator or a legal process may need.
7. Training (AI literacy)
- Article 4 of the EU AI Act requires deployers of AI systems to take measures on the AI literacy of their staff.
- Each person completes [training name] before they get access to an approved tool.
- Each person does a refresher every [12] months, and when we add a new tool.
- [Name] keeps a record of who completed the training and when.
8. Incidents and breaches
- If you put restricted data into an AI tool by mistake, tell [Name] within [1 working day]. Do not try to hide it.
- If personal data is involved, we follow our data breach procedure. A personal data breach can need a report to the Data Protection Commission within 72 hours.
- If an AI output caused harm to a customer, report it in the same way.
- A breach of this policy can lead to [disciplinary action under the staff handbook].
9. Review and acknowledgement
This policy was approved by [Name] on [date]. We review it every [12] months, and when the law or our tools change.
| Name | Role | Signature | Date |
|---|---|---|---|
| [Name 1] | [ ] | [ ] | [ ] |
| [Name 2] | [ ] | [ ] | [ ] |
| [Name 3] | [ ] | [ ] | [ ] |
How to fill it in: a 15-person retailer
Take a homewares retailer with two shops in Cork and an online store. Staff already use ChatGPT on their phones to write product descriptions and reply to reviews. The owner wants to keep the benefit and remove the risk.
- List what staff use today. The owner asks each team, without blame. The answers: ChatGPT on personal accounts, Copilot in Outlook, and an AI feature in the e-commerce platform.
- Choose the approved tools. The owner keeps Copilot (it is already in the company Microsoft 365 plan) and pays for a company ChatGPT workspace with training on business data switched off. Personal accounts stop for work.
- Adapt section 3. For a retailer the main restricted data is customer names with addresses, order history and card details. The owner adds supplier price lists to the confidential list.
- Set the output checks. Product descriptions need a check of dimensions, materials and care instructions against the supplier sheet. Replies to complaints need a manager’s review.
- Name the people. The operations manager owns the tool register and the training record. Incidents go to the owner by phone the same day.
- Run a 30-minute session. Each person reads the policy, sees two good and two bad examples, and signs section 9.
If the retailer later adds a WhatsApp agent for customers, that is a different project. It needs a spec, a test plan and probably a DPIA. The staff policy does not cover it.
Common mistakes in AI use policies
- A total ban. Staff will use the tools on their phones anyway, and you lose sight of it. Approve a safe option instead.
- No named tools. "Use approved AI tools" means nothing if nobody keeps the list. Section 2 must name the tool and the account type.
- Rules about the tool, not the data. The same tool can be safe on a company account with the right settings and unsafe on a personal account. Write rules about data and accounts.
- Forgetting AI features inside existing software. Meeting transcription, email drafting and document summaries are AI tools too.
- No route for mistakes. If staff fear punishment for an honest error, they will not report it, and you miss the 72-hour breach limit.
- Written once, never read. Tools change fast. Review the register at least every year and after each new tool.
How to roll the policy out
Send the policy with a short session, not as an email attachment alone. Show real examples from your own business: a good prompt with the names removed, and a bad one with a customer’s details in it. Record who attended in section 7.
Check the data terms of each approved tool before you add it to the register. The questions in our vendor questionnaire work for software tools too. If you build your own agent later, our GDPR guide covers the next steps.
Common questions
Can staff use the free version of ChatGPT for work?
Only for content with no personal or confidential data, if your policy allows it. A personal account runs under the user’s own settings and terms, which may let the provider use inputs to improve its models, and you cannot control or audit it. A company workspace or a tool in your existing Microsoft or Google plan gives you central settings and records.
Does the EU AI Act require an AI use policy?
No law requires a document with this name. Article 4 of the EU AI Act requires deployers to take measures that support the AI literacy of staff who use AI systems. A written policy plus a recorded training session is a simple way to show that you took such measures.
Who should own the policy in a small business?
One named person, often the operations manager or the owner. They keep the tool register, approve new tools, keep the training record and receive incident reports. If you have a DPO, they should review the data rules in section 3.
What if a staff member already pasted customer data into ChatGPT?
Treat it as a possible personal data breach. Find out what data went in and which account was used, delete the conversation where you can, and decide with your DPO or advisor whether the risk needs a report to the DPC. Then fix the cause, usually a missing approved tool.
Sources
- Regulation (EU) 2024/1689 (EU AI Act), Article 4, EUR-Lex
- AI Act Article 4 text and application date (AI Act Explorer, Future of Life Institute)
- Regulation (EU) 2016/679 (GDPR), Article 33, EUR-Lex
We last checked these sources on 19 November 2026. Rules, fees and prices change. Check the source before you act.
Bring us the task.
Tell us what happens today, which tools are involved and where it goes wrong. On a free 30-minute review, we’ll work out whether an agent, an automation or neither is the right fix, and what a sensible first release would be.
Start a project.

Book a call or send us a message. We read everything.
tash@neu.ie
Working with teams in Ireland and worldwide.
Free 30-minute AI review.