An AI acceptable use policy tells staff which AI tools they may use at work, what data they may put into them, what they must never use them for, and who checks the output before it reaches a customer or a decision. A useful one fits on two or three pages, names an owner, and is reviewed at least once a year. Below is a complete AI policy template you can copy and adapt.
Why you need an AI policy now
Your staff are already using AI tools, whether you have approved them or not. Someone in sales is drafting proposals in a chatbot. Someone in finance is pasting a spreadsheet into one to write a formula. Someone in customer service is using a browser extension that rewrites their replies. Most of this is sensible. Some of it puts customer data, contract terms or pricing into a service the company has no agreement with.
Banning AI rarely works; it just moves the use to personal phones, where you can see even less. A policy does three better things: it gives people a safe, approved way to use AI, it draws a clear line around the data that must not leave the company, and it makes someone responsible for checking what AI produces.
Before you write a word, ask. Short conversations with each team lead give you the outline: which tools the team uses and for what. Then ask two or three people doing the work, privately, what they paste into those tools on a busy day. The answers will differ. The lead may say the team only uses the approved assistant; the person answering customer emails may mention a free tool on their phone. That gap is exactly what your policy needs to close, so write it for the real use, not the official one.
What an AI usage policy must cover
- Approved tools. A short list of tools staff may use, with the account type (company account, not personal) and what each one is approved for.
- Data classes. Which kinds of information may go into which tools. Most companies need four classes: public, internal, confidential and personal data.
- Prohibited uses. Things AI must never be used for, regardless of tool.
- Human review. Who checks AI output, and which outputs must never go out unchecked.
- Being open about AI use. When customers, candidates or suppliers should be told that AI was involved.
- Accountability. The person who sent it owns it. AI is not an excuse.
- Training. What staff need to know before they use AI at work.
- Incident reporting. What to do when something goes wrong, such as confidential data pasted into the wrong tool.
- Review cycle. Who updates the policy and how often. AI tools change faster than most policies.
The data rule is the one that matters most
If staff remember one thing, it should be what they may paste where. A simple matrix works better than paragraphs of prose:
| Data class | Examples | Public AI tools (personal accounts) | Approved company AI tools |
|---|---|---|---|
| Public | Published prices, website copy, press releases | Allowed | Allowed |
| Internal | Process notes, internal emails, meeting notes without personal data | Not allowed | Allowed |
| Confidential | Contracts, margins, unreleased plans, supplier terms | Not allowed | Only tools approved for confidential data |
| Personal data | Customer names and contact details, employee records, CVs | Not allowed | Only where the tool is covered by a data processing agreement and the use has a lawful basis |
The difference between a personal account and a company account matters. Company or business plans of the major AI tools usually come with contractual terms about how your data is stored and whether it is used to train models; free personal accounts often don’t. Check each vendor’s current terms rather than assuming.
The regulation you should know about
This section is a summary for orientation. This is not legal advice; check the official sources and take advice for your situation.
EU AI Act
The EU’s Artificial Intelligence Act is Regulation (EU) 2024/1689. Its obligations apply in stages. Since 2 February 2025, Article 4 on AI literacy has applied to providers and deployers of AI systems: a company whose staff use AI systems at work must take measures to ensure those staff have a sufficient level of AI literacy, taking into account their role and the context the AI is used in. Training and a clear policy are the practical way most companies meet that. The bans on certain AI practices in Article 5 have applied from the same date.
In November 2025 the European Commission proposed changes to simplify parts of the AI Act and related digital rules (the “digital omnibus”). Proposals can change how and when some obligations apply, so check the current status on the European Commission’s website before relying on any date. The AI Act matters to companies outside the EU too if they place AI systems on the EU market or their AI output is used in the EU.
GDPR and UK GDPR
If staff put personal data into an AI tool, data protection law applies as it would to any other software: the EU General Data Protection Regulation (Regulation (EU) 2016/679) and, in the UK, the UK GDPR and the Data Protection Act 2018. In practice that means three checks for every tool that touches personal data:
- A data processing agreement with the vendor (GDPR Article 28 requires a contract with any processor handling personal data for you).
- Where the data is stored and processed, and whether transfers outside the EU or UK are covered.
- Whether the vendor uses your data to train its models, and whether you can switch that off.
The UK Information Commissioner’s Office (ICO) and EU data protection authorities publish guidance on AI and data protection that is worth reading alongside this.
AI acceptable use policy template
Copy this, replace everything in square brackets, and delete what doesn’t apply. Keep the wording plain: the people reading it are busy, and a policy nobody reads protects nobody.
1. Purpose
[Company] encourages staff to use AI tools where they help us serve customers and work better. This policy explains which tools you may use, what information you may put into them, and how to check what they produce. It applies to all employees, contractors and temporary staff who use AI tools for [Company] work, on any device.
2. What counts as an AI tool
Any software that generates text, images, audio, code or analysis from your input, including chat assistants, writing and meeting assistants, AI features inside tools we already use (email, spreadsheets, CRM), and browser extensions.
3. Approved tools
You may use only the tools listed below, through your company account. If you want to use a tool that is not listed, ask [role, e.g. IT lead] before you use it for work.
- [Tool name]: approved for [uses]. Data allowed: [public / internal / confidential].
- [Tool name]: approved for [uses]. Data allowed: [ ].
- AI features in [email / CRM / accounting system]: approved for [uses].
4. What information you may use
Public information may be used in any approved tool. Internal information may be used only in approved company tools. Confidential information (contracts, prices and margins, supplier terms, unreleased plans, anything under a non-disclosure agreement) may be used only in tools marked as approved for confidential data. Personal data about customers, employees or candidates may be used only in tools approved for personal data, and only for the purpose it was collected for. Never put passwords, bank details, or special category data (health, religion, ethnicity and similar) into any AI tool.
5. Prohibited uses
Do not use AI tools to:
- make or recommend decisions about hiring, pay, performance or dismissal without a person reviewing the case in full;
- make final decisions about customers’ credit, pricing or eligibility;
- create content that is misleading, discriminatory, harassing or unlawful;
- impersonate a real person or produce content that passes off AI output as their words;
- process data you are not authorised to access;
- [any use specific to your sector].
6. Checking AI output
AI tools make mistakes, including confident ones. You are responsible for anything you send, publish or act on, whether or not AI helped produce it. Before it leaves your hands, check facts, figures, names and references against the source. Anything sent to customers, published, used in a contract, or used to make a financial decision must be read in full by a person. [Role] must also approve [list, e.g. public statements, legal documents].
7. Being open about AI
If a customer or candidate is interacting with an AI system rather than a person, for example a chat assistant, tell them. Do not claim that content was written or checked by a person when it was not.
8. Training
Everyone who uses AI tools for work completes [Company]’s AI training before they start, and a refresher [annually]. The training covers how these tools work, their limits, this policy, and the data rules above. [Role] keeps a record of who has completed it.
9. Reporting a problem
If you put information into a tool it should not have gone into, or an AI output caused or nearly caused harm, tell [role / email address] the same day. Say which tool, what information, and when. Reporting quickly matters more than blame; nobody will be disciplined for reporting an honest mistake promptly.
10. Ownership and review
This policy is owned by [role]. Vendors and approved tools are reviewed by [role] before they are added. The policy is reviewed every [6 / 12] months, and whenever we add a tool or the law changes.
11. Breaches
Breaching this policy may lead to disciplinary action under [link to disciplinary policy].
Version: [x]. Approved by: [role]. Date: [date]. Next review: [date].
How to roll the policy out
- Start from real use. Collect which tools each team uses and for what, as described above. Approve the useful ones properly instead of pretending they don’t exist.
- Check each approved tool’s terms: data processing agreement, storage location, training on your data, and admin controls.
- Write the policy from the template, in the words your teams use. If customer service calls it “the assistant”, so should the policy.
- Train before you publish. A 30-minute session per team, using their own examples, does more than an email with an attachment.
- Keep it where people work: linked from the intranet, the tool’s login page or the team’s shared drive, not buried in an HR folder.
- Review it on a date, not when someone remembers.
When staff are unsure, give them one test: would you be comfortable if this text appeared on a supplier’s screen tomorrow? If not, it does not go into a tool that is not approved for confidential data. It is not a legal test, but it catches most of the mistakes.
Where the policy meets the work
A policy covers how individuals use general AI tools. Once AI is built into a process, such as reading supplier invoices or drafting replies to customer emails, the controls move into the process itself: who approves, what gets checked, and where the record lives. If you are at that stage, our AI readiness assessment helps you check whether a process is ready, and the lesson on going from a process map to automation shows how to keep a person approving each step that matters. For a worked example of automation with approvals kept in place, see the three-way match guide.
If you want an outside view of where AI could help your company first, the free Bonofici grader reads your public pages and shows where to start.