A company AI policy. One page of rules before someone pastes a contract into a chat
A short AI use policy fits on one page. Which tools we use, what never goes in, who checks the output. With a ready draft you can copy.
In this article
In short
An AI policy for a small business fits on one page and answers a few questions. Which tools and accounts we use, what never goes in, who checks the output and when we say that something was made with AI. Such rules also help with the duty to support AI literacy under Article 4 of the EU AI Act, and the data protection parts are worth checking with a lawyer.
A new person in the office gets a supplier contract and a question about the notice period. They paste the whole contract into a free chat on a private account and have the answer in a minute. Nobody told them not to, because nobody ever wrote it down.
A company AI policy does not have to be a thick rulebook. In a small business one page is enough, something everyone reads in five minutes and remembers.
Why a small business needs an AI policy
The first reason is practical. Without rules, everyone has their own. One person has switched off training on their conversations, another does not know the switch exists. One replaces names with labels, another pastes everything.
The contract from the opening is a good example. It holds the names of both firms, rates, deadlines and sometimes contact people's details. On a private account with training switched on, such text can be used to improve the model. Nobody did it out of bad will. There was simply no rule.
The second reason is the law. Article 4 of the EU AI Act, as worded by Regulation (EU) 2026/1744, says that firms using AI take measures to support the AI literacy of their staff. Short rules and one conversation about them are a natural start. Whether that is enough in your case is worth checking with a lawyer.
Eight points of an AI policy
This is what the table of contents can look like. Each point is two or three sentences.
- Which AI tools we use at work.
- On which accounts, private or company.
- What we never paste in.
- How we hide client data before pasting anything.
- Who checks the output before it reaches a client.
- When we say that something was made with AI.
- Who we report a mistake or slip to.
- When we review these rules.
Tools and accounts
Write down the names of the tools you use, not a general "AI tools". If someone wants to try a new one, they ask the person who keeps the list.
Accounts are best set up on a company email address. When someone leaves, the chat history and settings stay with the firm. Business plans have one more difference. By default OpenAI does not use content from ChatGPT Business and Enterprise for training, and Anthropic does not use conversations from its Team and Enterprise plans. On private accounts you have to switch training off yourself, as we described in Client data in ChatGPT.
What never goes in, and who checks
Passwords, national ID numbers, passport and card numbers never go into a chat. Once pasted, they sit on someone else's server and no setting can undo that. Names, phone numbers and invoice numbers become "Client A" and "Invoice 1". The answer is just as good.
Whether your use of AI needs a data processing agreement or a change in your records of processing is a GDPR question for a lawyer.
A person checks the output before it reaches a client. AI writes in a confident tone even when it is wrong, so someone always reads the numbers, dates and names.
Labelling and mistakes
Since 2 August 2026 the EU AI Act requires you to disclose when you publish AI generated images, video or audio that could pass as real. The same goes for text that informs the public on matters of public interest, unless it went through human editing and someone holds editorial responsibility.
Beyond that duty, decide yourselves when you tell clients about AI. What matters is that everyone in the firm does it the same way.
A mistake reported at once costs less than a hidden one. Write down the name of one person to go to.
A draft to copy
# How we use AI
Tools. ChatGPT on the company account. A new tool only after talking to Anna.
Accounts. Company email only. Training on conversations switched off.
Never paste. Passwords, ID numbers, passport and card numbers.
Client data. Names and phone numbers become Client A.
Checking. A person reads numbers, dates and names before sending.
Labelling. AI images and video are published with a note that they are AI.
Mistakes. Report them to Anna at once.
Review. Once a quarter.The name is made up, put your own. Print this page, read it with your people and come back to it in a quarter.
At the first meeting, start by asking who already uses what. The list usually turns out longer than expected, because someone uses an AI translator and someone else makes images for posts. It is better to know and put it on the list than to pretend there is no AI in the firm. Rules nobody follows do nothing, so write only the ones you really mean to keep.
If you would rather start with where AI makes sense for you at all, book a free AI audit. After a conversation of about 30 minutes you get a written list of 3–5 places where AI can save time, with the tool for each and what to watch out for with client data. No obligation.
Documentation for this topic
What would you like to change in your business?
Leave your number. We will talk about your process and see whether we can help. The first call is free.
The service this text is about
Automation and AI / OCRAutomation of document re-typing, OCR and checking rules, and your data stays with you.