Source. This piece is adapted from the AI acceptable use policy template by Brannan Coady of Anyway Ventures, version 1.0, October 2026. Anyway Ventures publishes it as free to use, adapt and share with attribution. It is general guidance and is not legal advice.
No UK law requires a document called an AI policy. Plenty of law already applies to how staff use AI tools. UK GDPR covers any personal data pasted into a tool. The Equality Act 2010 covers decisions about people. Client contracts and professional rules cover confidential material.
The template's starting point is that a short policy, a list of approved tools and a training record are the simplest way to show a business has taken appropriate measures. It is written for owners, COOs, HR directors, data protection leads and heads of IT in businesses of roughly 20 to 2,000 staff.
The time it asks for is modest. A small drafting group needs five working days to draft and agree the policy. Rolling it out, training people and checking adoption takes four to eight weeks.
Approve a tool before you write the rules
The template cites Microsoft's UK research, run by Censuswide with 2,003 employees in October 2025. It found that 71% had used unapproved consumer AI tools at work, and 28% said their employer offered no approved option.
A policy that only prohibits moves that use onto personal phones. On a personal phone the business has no contract terms with the provider, no logs and no say over what happens to the data.
The order matters for a second reason. The data rules depend on the terms of the tool. A business plan with data processing terms is materially different from a consumer account, so approving a tool after the policy is written means rewriting the data rules.
Clause 3 of the template therefore points to an approved tools register. For each tool, the register records the plan, who may use it and the most sensitive category of data it is approved for. The drafting note on that clause puts the point in one line: "Use AI responsibly" is not a rule; "use Copilot through your work login" is.
The traffic-light data table
"Do not input confidential information" leaves every employee to decide what confidential means. The template replaces that judgement with a green, amber and red table that staff can check in seconds.
| Colour | What it includes | Where it can go |
|---|---|---|
| Green | Public information, published marketing, general knowledge questions, your own writing with no names or identifiers, and properly anonymised examples. | Any approved tool, including personal and free accounts. |
| Amber | Internal documents and reports, non-sensitive business data, supplier and pricing information, customer names and business contact details, and meeting notes that are not red. | Approved business accounts only, within the level each tool is approved for. |
| Red | Special category data such as health, ethnicity and religion. Criminal records. HR, disciplinary, grievance and appraisal records. Payroll and bank details. Passwords, keys and credentials. Unreleased financial results and inside information. Anything an NDA or contract restricts. Children's personal data. | No AI tool, unless the policy owner and the data protection lead have approved that specific use in writing, in a specific tool. |
The template then gives six decision rules for the cases the table does not settle.
- If you are not sure of the colour, treat it as red and ask.
- The highest colour wins. A spreadsheet with one red column is red.
- Combinations can change the colour. Two amber files joined together can identify people or reveal something neither does alone.
- Making data greener is allowed. Remove names, replace them with roles such as "Customer A", round the figures and strip identifiers. Pseudonymised data is still personal data if the key exists.
- The tool's level and the data's colour must both allow it. Amber data in a tool approved only for green is a breach, even if the tool is a business account.
- Output inherits the colour of the input. A summary of an amber document is amber.
The template's advice is to replace its examples with ones from your own teams, because people learn the rule from the example. In finance, explaining an accounting standard is green, the management accounts are amber and payroll is red. In HR, drafting a generic job advert is green, an organisation chart is amber, and CVs and interview notes are red.
What the fifteen clauses cover
The policy itself is fifteen short clauses. Each one comes with a drafting note that explains why it is there and how to make it lighter for a small firm or stricter for a regulated one.
- Purpose. The business wants people to use AI to do better work faster. A policy that opens with risk gets read as a ban, and use goes underground.
- Scope and definitions. Everyone working on the company's behalf, on any device, including AI features built into software the business already uses.
- Approved tools. Use the tools on the register, through accounts the company provides.
- Data rules. Check the colour of the information and the approved level of the tool before putting anything in.
- AI meeting assistants and recording. Tell everyone before a tool records or transcribes, and keep recording out of HR, disciplinary, grievance, medical and legal meetings.
- AI agents and connections to company systems. IT approves every connection. Every agent has a named owner, its own login and the least access it needs.
- Prohibited uses. A short list of fixed prohibitions, led by using AI to make final decisions about recruitment, pay, performance, discipline or dismissal.
- Human review. You are responsible for work you produce with AI. A named person reviews output before it goes to a customer, into a financial figure or into a decision about a person.
- Transparency and disclosure. Tell your manager about significant AI-assisted work, and tell customers when their contract requires it or they would reasonably expect to know.
- Intellectual property and confidentiality. Work created with AI in the course of the job belongs to the company. Check output for copied material before publishing or shipping it.
- New tools and procurement. A request form, a target decision time of ten working days, and a 30-day amnesty for tools already in use.
- Security and impersonation. Verify any request to change bank details or make an urgent payment by calling back on a number already held.
- Training. Everyone with access to an approved tool completes training within 30 days and refreshes it every 12 months.
- Incidents. Report mistakes within 24 hours. Reporting quickly will not, by itself, lead to disciplinary action.
- Ownership and review. A named owner, an approval date and a review date no more than six months later.
The periods in clauses 11, 13 and 14 are the template's suggested defaults, and each business can change them.
The template's tip on length is to delete before you add. A 25-person firm with no agents and no client meetings can remove clauses 5 and 6 entirely. Every clause kept is a rule someone has to follow and someone has to enforce.
Three risks most templates miss
The template argues that most AI policies leave out three risks, and it gives each one its own clause.
AI note-takers. They are built into most video platforms and are often switched on by default. A transcript of an appraisal or a client negotiation is personal and often confidential data, and an AI summary can attribute words to the wrong person. Clause 5 requires notice to everyone in the meeting and treats transcripts as at least amber data.
AI agents. An agent that can read incoming email can be instructed by whoever writes the email, which is known as prompt injection. An agent using your credentials can do anything you can. Clause 6.5 stops an agent holding three things at once without IT's assessment: access to confidential information, content from outside the business, and the ability to send messages externally.
Voice and video impersonation. The template cites the 2024 case in which the engineering firm Arup confirmed an employee in Hong Kong had sent about US$25 million after a video call with what appeared to be the chief financial officer and colleagues, according to Hong Kong police. The control in clause 12 is procedural. Call the person back on a number the business already holds, however senior they appear to be.
People decide about people
Since 5 February 2026 the Data (Use and Access) Act 2025 has replaced Article 22 of UK GDPR with Articles 22A to 22D. The new articles allow more solely automated decisions, with safeguards.
The template notes that the Information Commissioner's Office (ICO) consulted on draft guidance on automated decision-making between 31 March and 29 May 2026. The draft, and the ICO's Recruitment rewired report, make clear that a rubber-stamp review does not count as human involvement.
For most businesses the template recommends a simple rule. AI may draft and sort, and a trained person decides on hiring, pay, performance, discipline and credit.
Clause 8.5 defines what a review has to involve. The reviewer has the knowledge and the authority to change the result. The reviewer has looked at the evidence and not only the AI's summary. The review happens before the output is used.
The law behind the clauses
The UK has no AI Act. What applies is the law a business already follows, applied to a new kind of tool. The template maps each law to the clauses that respond to it, and states that its legal summary was checked to 25 October 2026.
- UK GDPR and the Data Protection Act 2018 apply whenever personal data goes into or comes out of an AI tool. A personal data breach likely to result in a risk to people must be reported to the ICO within 72 hours of the business becoming aware of it.
- The Data (Use and Access) Act 2025 applies when AI is used in a significant decision about a person. From 19 June 2026 it also requires every controller to acknowledge data protection complaints within 30 days.
- The Equality Act 2010 applies whatever tool produced a decision. An AI screening tool that disadvantages candidates who share a protected characteristic can amount to indirect discrimination.
- Client contracts, NDAs and professional duties make no exception for AI tools. Putting restricted information into a tool whose provider retains it can breach those terms.
- Copyright law still applies to output. AI output can infringe if it reproduces a substantial part of someone else's work.
- The Digital Markets, Competition and Consumers Act 2024 has banned fake consumer reviews since 6 April 2025, including ones written with AI.
- The EU AI Act is most likely to catch a UK business that sells an AI-enabled product to EU customers, has an EU subsidiary or office, or uses AI to produce work that is delivered to and relied on in the EU. A purely domestic business is outside it.
The pack also carries add-on clauses for financial services, legal services, healthcare, education and recruitment. For firms regulated by the Financial Conduct Authority (FCA), it notes that the FCA has said it does not plan extra regulations for AI. An AI failure is judged against rules the firm already follows, particularly the Consumer Duty and the Senior Managers and Certification Regime.
Rolling it out
The template treats adoption as a dated plan with named owners.
- Week 0. The policy is approved, the tools register is completed and business accounts are set up. The CEO announces why the policy exists, what is approved, the 30-day amnesty and the training dates.
- Week 1. The policy, the register, the staff summary and the forms are published internally.
- Weeks 1 to 4. All users are trained, by team, on their own work. Managers are briefed.
- Week 4. The amnesty closes, and the tools staff disclosed are reviewed through the request route.
- Week 6. The anonymous usage survey is repeated and compared with the baseline.
- Week 8. A short report goes to the CEO on training completion, requests, incidents and the change in the survey.
- Month 6. The first formal review.
Before approval, the template suggests a read-through by four or five staff from different teams. Ask them one question: is there anything here you could not follow tomorrow?
The amnesty in clause 11.4 is the fastest way to find out what is in use. Anyone already using a tool that is not on the approved list tells the policy owner within 30 days and is not disciplined for it, unless red data was involved.
Training is where the template says a policy turns into behaviour. It cites the Office for National Statistics, which reported in July 2026 that around 35% of UK businesses with ten or more employees use AI. Only 11% say more than half their workforce has had AI-related training.
The approach it recommends is teaching the rules while people use the tools on their own tasks, so that checking the colour becomes a habit. The training record does a second job for a business that sells into the EU. It is the evidence for the AI literacy duty in Article 4 of the EU AI Act.
What staff need to remember
The pack includes a one-page staff summary to hand out alongside the full policy. These eight rules are what people should be able to recite after training.
- Use the approved tools, through your work account. Personal accounts are for green data only.
- Check the colour before you paste. Green goes anywhere approved. Amber goes into business accounts only. Red does not go in without written approval. If you are not sure, it is red.
- You own the output. Check facts, figures and sources before anything leaves the building or goes into a number.
- People decide about people. AI can help draft, and never decides, on hiring, pay, performance or discipline.
- Say so when it matters. Tell your manager about significant AI-assisted work, and the customer when they would expect to know.
- Ask before you buy or connect. New tools, note-takers and connections to company systems go through the policy owner.
- Check before you pay. A familiar voice on a call is not proof. Call back on a number you already hold.
- Report mistakes fast. Within 24 hours, to the policy owner. Owning up quickly is treated well.
Common mistakes
The template lists twelve common mistakes. Six of them are not covered by the sections so far.
- Forgetting AI inside existing software. Vendors switch AI features on in CRM, HR, finance and video software with routine updates. Ask IT to check settings quarterly and put each feature in the register.
- Letting agents borrow personal credentials. An agent running as you can do anything you can, and its actions look like yours in the logs.
- A slow request route. If approval takes months, people will not wait. Set a target of ten working days and publish the decisions.
- Treating the policy as the training. Asking staff to read and sign a policy produces signatures. Teaching the rules on their own work changes what they do.
- Copying a US template. US templates refer to state laws and at-will employment, and they miss UK GDPR, the Equality Act and the ICO.
- Over-reading the EU AI Act. A UK-only business does not need an AI Act compliance programme. Article 4 asks for measures and a record, and no certificate is needed.
Where to get the full template
This write-up leaves out most of the pack. The full template from Anyway Ventures runs to 35 pages. It has the complete policy wording, a drafting note on every clause, the approved tools register, a use-case request form, an incident report form, a manager's FAQ, the five sector add-ons, a worked example, a glossary and the list of sources.
Attribution. The template was written by Anyway Ventures, the York technology and AI practice led by Brannan Coady. It is version 1.0, dated October 2026, with law and guidance checked to 25 October 2026. The policy content, the data table, the statistics and the legal summary in this piece are theirs. Any errors in summarising them are ours. This piece is general guidance and is not legal advice.