Getting started

How to write a staff AI policy in 5 steps

A staff AI policy fits on one page. Find out what staff use, name the allowed plans, list what may go in, ask for a check and name an approver.

A staff AI policy fits on one page, and you can write it in 5 steps. Find out which AI your staff already use. Name the plans they may use for company work. List what they may put in and what they must never put in. Ask them to check every answer. Name the person who approves new tools.

A staff AI policy is also called an AI policy for employees, or an AI usage policy. Our free staff AI policy tool builds the page as you tick boxes, and this guide explains each choice.

Step 1: find out which AI your staff use

Ask the team, without blame. Most owners we talk to do not know which AI plan their staff use. Write down each tool and whether the account is personal or paid for by the company.

The plan matters more than the brand. A personal plan can learn from what staff type, and no contract covers your data. Our safe AI check shows what each plan does with company data.

Step 2: name the plans staff may use

List the plans staff may use for company work, by name. Add one rule. Staff always sign in with the company account.

For personal accounts, the safe rule is "not for company work, even on a paid plan". If you allow personal plans at all, limit them to public information and ask staff to turn training off on each account.

If you are choosing a plan now, these compare the common ones:

Free apps such as DeepSeek need a rule too. Is DeepSeek safe for company work? explains why.

Step 3: list what may go in, and what never does

Go through the kinds of information your company holds, such as customer details, staff records, prices and contracts. For each kind, decide whether staff may put it into the allowed AI.

Passwords, NRIC numbers and bank details go on the "never put in" list in every company.

Where staff may put in personal data, they put in only the parts the task needs. To summarise a complaint, the AI needs the complaint. It does not need the customer's phone number.

Anything your policy does not allow goes on the "never put in" list. A short list of what is allowed is easier to follow than a long list of what is banned.

Step 4: ask staff to check every answer

AI writes confident answers that are sometimes wrong. The policy asks staff to read every AI answer before it goes outside the company, and to check names, numbers and dates against the source.

Add one small rule if you allow the ChatGPT personal plans, Claude Team or Claude Enterprise. Pressing the thumbs up or thumbs down button can send the whole chat to the AI company. Staff should leave those buttons alone in any chat that holds company information. Our safe AI check lists the plans this applies to. OpenAI's data controls page and Anthropic's training page give the details.

Step 5: name who approves new tools, and set a review date

New AI tools keep appearing. Name one person who approves a tool before staff use it for company work. The same person reviews the policy every six months, because providers change their terms often.

What the regulator expects

Singapore's data protection regulator, the PDPC, published guidelines on personal data in generative AI. The company is responsible for checking the AI tool it picks. The company should tell staff which personal data they may put in, and the PDPC encourages a written policy. More details.

This guide is a starting point and not legal advice. Read your policy through and change anything that does not fit your company.

Questions people also ask

What is an AI usage policy for employees?

It is a short written set of rules for how staff use AI at work. It names the AI plans they may use and what they may put in. The PDPC's guidelines encourage a written policy and make the company responsible for checking the AI tool it picks. More details.

Is there a free AI policy template for employees?

Yes. Our staff AI policy tool is free. You tick what applies to your company, and it writes a one-page policy you can copy or save as a PDF. The Word file asks for your email.

Can employees use ChatGPT at work?

Yes, on a business plan with a company account. A personal account is not for company work, even on a paid plan. A personal plan can learn from your chats, and no contract covers your data.

How long should an AI policy be?

One page is enough for a small company. The page covers the allowed plans, what may go in, the check before anything goes out, and who approves new tools.

Where to start

This week, ask your team which AI tools they use and on which accounts. Then open the staff AI policy tool, tick what applies and send the page to your staff.

To see an AI agent that works inside these rules, book a call.

Keep reading

More from the lab

All articles