
A team that is told to start using AI will start, whether or not anyone has said how. People paste meeting notes into free tools, draft client emails with a chatbot and upload spreadsheets to get a summary. Most of this is harmless. Some of it puts customer data, contracts or unreleased plans into a tool your company has never checked. A short AI use policy fixes that before training starts, and it makes the training easier to run because everyone begins with the same rules.
This guide sets out what a staff AI use policy should cover in Singapore, in plain language, and where Singapore’s published guidance fits. It is practical guidance from a training provider and is not legal advice. For the legal position on personal data, take advice and read the PDPC’s own pages.
Why write the policy before the training
Training shows people what AI tools can do. A policy tells them where the limits are. If you train first, staff leave motivated and put the new skills to work on whatever data is closest. If the rules come first, the training can use them: learners practise on approved tools and safe data, and they leave knowing what is allowed. Our guide to planning a corporate AI training programme places the policy at the start of the plan for this reason.
What a staff AI use policy should cover
Which tools are approved
List the tools staff may use for work and say how a new one gets added. Free consumer accounts and company-managed accounts often handle data differently, so name the account type as well as the product. For a small business, a simple request route is enough, such as asking the person who looks after IT or operations.
What data can go in
This section matters most. Say plainly what may not be entered into an AI tool: customer personal data such as names, contact details and identity numbers, confidential contracts, passwords and unreleased financial results. Give a safe alternative, such as removing identifying details first. Our PDPA checklist for using AI on customer data walks through that habit step by step.
Who checks the output
AI tools make mistakes and sound confident while doing it. State that a person reviews anything that goes to a customer, a regulator or the public, and that the person who sends it is responsible for it, whichever tool helped draft it.
When to say AI was used
Decide when AI use must be disclosed, for example in client deliverables or published content, and apply the rule evenly. Staff follow a specific disclosure rule more easily than a general instruction to be transparent.
What to do when something goes wrong
Name a person or a mailbox, and tell staff to report a data leak, a wrong output that was already sent or a suspicious tool straight away and without blame. A company finds problems faster when reporting is easy.
How the policy gets updated
Tools change quickly. Give the policy an owner and a review date so it stays accurate.
Where Singapore guidance fits
Several published documents can sit behind your policy. The PDPC’s Advisory Guidelines on Use of Personal Data in AI Recommendation and Decision Systems give organisations more clarity on using personal data to develop AI, on the information to give consumers when seeking consent, and on the obligations of third-party developers of bespoke AI systems.
The PDPC has also issued Advisory Guidelines on Use of Personal Data in Generative AI. They cover how personal data can be collected and used to develop generative AI models, how data protection responsibilities are shared across the generative AI lifecycle, and how requests from individuals about their personal data should be handled. Both sets of guidelines are advisory and are meant to be read with the PDPC’s Advisory Guidelines on Key Concepts in the PDPA.
For a wider view of good practice, the Model AI Governance Framework for Generative AI, developed by the AI Verify Foundation and IMDA, names nine dimensions: accountability, data, trusted development and deployment, incident reporting, testing and assurance, security, content provenance, safety and alignment research and development, and AI for public good. It is written mainly for AI developers and policymakers rather than for a business buying tools, but several dimensions map directly onto a staff policy. Accountability maps to who owns decisions, data to what staff may enter, incident reporting to your reporting route and content provenance to your disclosure rule.
Keep the policy short
A policy that fits on two pages gets read. Use plain sentences, give examples of allowed and disallowed use, and put it where staff already look. Pair it with training so people can practise the rules on real tasks.
How training supports the policy
A policy on paper does not change habits. Hands-on training does, because learners practise choosing a tool, preparing data and checking output. In our one-day AI-Powered Productivity course, learners compare the major AI tools and build an agent on a process from their own role, which is a natural point to apply your data rules. If you are planning for a team, our guide to choosing between an in-house workshop and a public course helps you decide the format.
Your next step
If you want training for a team and have not written a policy yet, start with the six sections above and bring the draft to the planning conversation. Message our team on WhatsApp at +65 8069 3194 and tell us the roles involved and the tasks you want to change. We will suggest a format and check which funding schemes apply, and you can work out a first estimate with the grant calculator on our Grants and Subsidies page. Our guide to funding AI training for your team explains each scheme.
Frequently Asked Questions
Does my company need an AI use policy?
A short policy is sensible for any team that uses AI at work, because without one each person decides alone what is safe. It tells staff which tools and data are allowed, who checks output and how to report problems.
What should a staff AI use policy include?
Approved tools, what data may be entered, who reviews output, when AI use is disclosed, how to report incidents and who keeps the policy up to date. Two pages is enough for most small teams.
Can staff put customer data into ChatGPT or other AI tools?
Treat customer personal data with care. Remove names, contact details and identity numbers before using any AI tool, and use only tools your company has approved. The PDPA applies to personal data whichever tool is used, and this is not legal advice, so check with your data protection officer or a lawyer for your situation.
Where can I find official guidance on AI and personal data in Singapore?
The PDPC publishes advisory guidelines on the use of personal data in AI recommendation and decision systems and in generative AI on its regulatory guidance pages. IMDA and the AI Verify Foundation publish the Model AI Governance Framework for Generative AI.
Should the AI policy come before AI training?
Yes, where you can. Staff then practise the rules during the training, using approved tools and safe data.