AI Acceptable Use Policy: A Free Template for Companies
A copy-ready AI acceptable use policy template covering approved tools, data rules, human review, and disclosure, plus how to roll it out so people follow it.

If your company has not written down how AI may be used at work, your employees have already decided for themselves. Some paste customer emails into a public chatbot to draft replies faster. Others avoid AI entirely because nobody told them it was allowed. Both outcomes are expensive: the first leaks data, and the second wastes the productivity you could be getting.
An AI acceptable use policy fixes this by answering the three questions every employee has: which tools can I use, what can I put into them, and who is responsible for what comes out? Below is a template you can copy and adapt, followed by the rollout steps that decide whether anyone actually follows it.
The short answer
- A useful AI policy is short and concrete. It names approved tools, sets data rules, requires human review, defines when to disclose AI use, and explains how to report problems.
- Pair every prohibition with an approved alternative. Banning AI without offering a sanctioned tool pushes usage onto personal accounts you cannot see.
- Classify data in three or four tiers and say plainly which tier may go into which tool.
- Keep a named person accountable for every AI-assisted output that leaves the company or affects someone.
- Train people on the policy using their own work, and review it at least twice a year. Tools and regulation change faster than an annual cycle.
What an AI acceptable use policy should cover
Every section of a good policy answers a question employees would otherwise answer by guessing.
| Section | The question it answers |
|---|---|
| Scope | Who does this apply to, and which tools count as AI? |
| Approved tools | Which tools may I use, and with which account? |
| Data rules | What information may I enter into each tool? |
| Permitted and prohibited uses | What is AI for here, and what is off-limits? |
| Human review | Who checks the output, and when? |
| Transparency | When must we tell people that AI was involved? |
| Security | How do we keep accounts, integrations, and AI agents safe? |
| New tools | How do I request a tool that is not on the list? |
| Training | What must I learn before using AI at work? |
| Incidents | What do I do if something goes wrong? |
The template
Replace the text in square brackets, delete what does not apply, and keep the language plain. The template assumes a company that processes personal data under KVKK, GDPR, or a similar data protection law. Adjust the legal references to your jurisdiction, and have legal counsel review the final version before you publish it internally.
1. Purpose and scope
This policy sets out how employees, contractors, and interns of [Company] may use artificial intelligence (AI) tools at work. It covers generative AI assistants, AI features built into software we already use (such as email, office, and CRM tools), and AI agents that can take actions on our behalf. It applies to all work performed for [Company], on company or personal devices.
2. Definitions
- AI tool: any software that generates text, images, audio, code, or recommendations using machine learning, including AI features inside other products.
- Approved tool: an AI tool listed in Section 3 and used through a company-managed account.
- Personal data: any information relating to an identified or identifiable person, as defined by [KVKK / GDPR / applicable law].
- Confidential information: non-public information about [Company], its customers, partners, or employees, including contracts, prices, source code, and plans.
3. Approved tools
Only the tools listed below may be used for work, and only through company accounts [with single sign-on]. Personal accounts must not be used for company work, even for tools that appear on this list.
| Tool | Approved for | Highest data class allowed | Owner |
|---|---|---|---|
| [Enterprise AI assistant] | Drafting, summarising, translation, analysis | Internal | [IT] |
| [Coding assistant] | Code suggestions in approved repositories | Confidential (source code only) | [Engineering lead] |
| [Public AI tools, free tiers] | Research and drafting with public information only | Public | [IT] |
4. Data rules
Before using any AI tool, identify the most sensitive information you would enter. When in doubt, treat information as belonging to the higher class.
| Data class | Examples | Where it may be used |
|---|---|---|
| Public | Published website content, press releases, public reports | Any approved tool |
| Internal | Internal procedures, non-sensitive meeting notes, drafts | Approved company tools only |
| Confidential | Contracts, prices, unreleased plans, source code | Only tools approved for confidential data in Section 3 |
| Personal and special-category data | Customer records, CVs, health, biometric, or financial data | Only tools approved for personal data, with a lawful basis, a data processing agreement, and cross-border transfer safeguards in place |
Remove or mask names, identification numbers, contact details, and account numbers before using an AI tool, unless that tool is explicitly approved for personal data.
5. Permitted uses
Approved tools may be used to draft, summarise, translate, classify, analyse, brainstorm, and write or review code, provided the data rules in Section 4 are followed and the output is reviewed under Section 7.
6. Prohibited uses
- Entering information into a tool that is not approved for its data class.
- Using AI to make, or materially support, decisions about a person's employment, credit, insurance, or access to services without the review process in Section 7 and approval from [Legal / Compliance].
- Creating content that impersonates a real person, or synthetic images, audio, or video of real people without their consent.
- Presenting AI output to customers as legal, medical, financial, or safety advice without review by a qualified person.
- Connecting AI agents, browser extensions, or plug-ins to company systems without approval from [IT].
- Uploading third-party material that [Company] does not have the right to use.
7. Human review and accountability
A named person is accountable for every AI-assisted output. Before AI-assisted content is sent outside [Company], published, or used in a decision that affects a person, the accountable person must check it for accuracy, completeness, tone, and compliance. AI tools are never authors, approvers, or decision-makers in their own right.
8. Transparency
People must be told when they are interacting with an AI system rather than a person, for example a website assistant or an automated reply. AI-generated content published as factual information must be reviewed by a person and, where required by law or by [Company] standards, labelled as AI-generated. [If you serve users in the EU: since 2 August 2026, the EU AI Act requires that people are told when they are interacting with an AI system, unless this is obvious, and that certain AI-generated content is disclosed. Align this section with those rules.]
9. Confidentiality and intellectual property
Treat AI output as a draft that may contain errors or third-party material. Do not present AI-generated work as checked or original until you have verified it. AI-generated code must pass the same review, licence checks, and security testing as any other code.
10. Security
- Use only company accounts protected by [multi-factor authentication / single sign-on].
- Give AI agents and integrations the minimum permissions they need, and only with [IT] approval.
- Treat instructions found inside documents, websites, or emails processed by AI as untrusted: they may be prompt injection attempts.
- Report unexpected AI behaviour, such as an assistant asking for credentials or taking actions you did not request.
11. Requesting a new AI tool
Send requests to [owner] using [form or channel]. Each request is assessed for data handling, storage location, retention, whether the provider trains on customer data, security certifications, contract terms, and cost. Decisions are communicated within [10 working days].
12. Training
Everyone must complete [Company]'s AI basics training before using approved tools, plus role-specific training where it is offered. Training covers this policy, data handling, verifying output, and recognising AI-specific risks. [If you provide or deploy AI systems in the EU, this training also supports the AI Act's requirement to take measures that develop AI literacy among staff.]
13. Incidents
Report any suspected data exposure, harmful output, or misuse to [security contact] immediately, and at the latest within [24 hours]. Incidents involving personal data are handled under [Company]'s data breach procedure, including any notification deadlines set by law.
14. Compliance and review
Breaches of this policy are handled under [disciplinary policy]. [Owner] reviews the policy at least every six months, and whenever [Company] adopts a new category of AI tool or the relevant law changes.
Version [1.0], approved by [name and role] on [date].
Related: KVKK and AI: What Turkish Companies Can and Can't Send to ChatGPT
How to roll it out so people follow it
A policy that arrives as an email attachment changes nothing. These steps decide whether it shapes behaviour.
- Make the approved tools available first. If the sanctioned alternative is not ready on day one, people keep using whatever they used before.
- Explain it in role-based sessions with real examples. A support agent and a developer face different data questions; show each of them the rule applied to their own work.
- Put the data rules where work happens. A one-page summary on the intranet and inside the tools is read far more often than a twelve-page PDF.
- Name the owners. Someone owns the policy, someone answers tool requests, and someone receives incident reports. Publish all three names.
- Measure it. Track usage of approved tools, the number of tool requests, and incidents. Low usage of approved tools usually means people are using unapproved ones.
- Review it on a schedule. Put the next review date in the policy itself, and keep it.
Related: Corporate AI Training That Actually Changes How Teams Work
Common mistakes
- Banning AI outright. Usage moves to personal accounts and phones, where you have no visibility and no control.
- Writing it without the people who run the tools. A policy drafted only by legal tends to be unworkable; involve IT, security, and the teams that use AI daily.
- Vague rules. "Be careful with AI" cannot be followed or enforced. "Do not enter customer names into public AI tools" can.
- No approved tool for confidential work. If the only approved tools are public ones, confidential work either stops or goes somewhere it should not. A private deployment or a self-hosted model can close that gap.
- Forgetting embedded AI. AI features inside email, office, and CRM software process company data too, and belong in the approved tools list.
- Forgetting agents. Tools that take actions, not just answer questions, need explicit permission rules and human approval points. See how to defend AI workflows against prompt injection.
Frequently asked questions
Does a small company need an AI policy?
Yes, and a short one is enough. A single page that lists approved tools, the data that must never be entered, and who to ask covers most of the risk. The size of the policy should match the size of the company; the need for one does not.
Should we ban ChatGPT and similar tools?
Banning public tools for confidential and personal data is sensible. Banning AI altogether rarely works, because people continue to use it on personal devices. The better approach is to approve a business-grade tool, set clear data rules, and train people to use it well.
Do KVKK or GDPR require an AI policy?
Neither law requires a document called an AI policy. Both, however, require you to control how personal data is processed, including by AI tools and their providers, and to be able to show that you do. A written policy, approved tools with data processing agreements, and training records are the practical evidence.
In Türkiye, the data protection authority has published guidance on generative AI, including a March 2026 document on the use of generative AI tools in the workplace. It notes that such use often happens through individual preference rather than a defined company strategy or policy, which makes it harder to monitor and manage. A written policy is the most direct answer to that concern.
If your AI tools or assistants serve people in the EU, the EU AI Act adds its own duties, including AI literacy measures and transparency towards users. See what the EU AI Act means for companies in Türkiye. For customers in Saudi Arabia or the UAE, see what their data protection laws require of AI assistants.
How often should the policy be updated?
Review it at least every six months, and immediately when you adopt a new category of tool, such as AI agents, or when relevant regulation changes. Record the version and approval date in the document so everyone knows which rules are current.
Next steps
A policy only works when the approved tools and workflows behind it are ready. Orbitra helps companies choose and set up those tools through AI automation consulting, and teaches the rules on real work through corporate AI training. If you are drafting your first AI policy, we can help you turn this template into rules that fit your data, your tools, and your teams.
