Your team is probably already using AI tools. The risk is rarely the tool itself. It is what people paste into it: customer data, contracts, passwords, HR files. An AI acceptable-use policy gives everyone the same simple rules. It also supports your duties under the GDPR and the EU AI Act.
Why you need one
- GDPR: personal data in a prompt is processing. You need a legal basis, a vendor agreement and a decision about where data goes.
- AI literacy: the AI Act expects providers and deployers to take measures that support AI literacy among staff who use AI.
- Transparency: Article 50 can require telling people when they interact with AI or see AI-generated media. See our Article 50 guide.
What to include: 8 sections
- Purpose and scope. Who the policy covers: employees, contractors, freelancers.
- Approved tools. A named list. Anything else needs approval first.
- Never-enter list. Customer personal data, employee data, special-category data, passwords, confidential contracts, source code.
- Account rules. Use company accounts with a data processing agreement, not personal free accounts. Check whether your data is used to train models and switch that off where possible.
- Human review. A person checks AI output before it goes to a customer, is published, or affects a decision about a person.
- Transparency. When people must be told they are dealing with AI, and how AI-generated images, audio and video are labelled.
- Prohibited uses. No emotion recognition on staff, no social scoring, no automated decisions on hiring or firing without review.
- Incidents and training. Report mistakes immediately, because a leak to an AI tool can be a personal data breach. Train staff at onboarding and yearly.
Sample clauses you can adapt
- "Only the AI tools on the approved list may be used for work. New tools require approval from [name or role]."
- "Never enter customer or employee personal data, credentials or confidential contracts into an AI tool unless it is approved for that data type."
- "You remain responsible for anything you send or publish. A person must review AI output before it is shared with customers."
- "If you enter sensitive data into an AI tool by mistake, tell [contact] immediately."
Fill a short form and download your AI acceptable-use policy as Word or PDF.
Rolling it out
- Take stock: ask the team which AI tools they already use, with no blame.
- Approve what is safe and replace what is not.
- Publish the policy, walk through it in a 15-minute meeting and record attendance.
- Review it every six months, because tools and rules change.
Record AI tools that handle personal data in your record of processing, and screen higher-risk uses with a DPIA. The GDPR checklist covers the rest.
Frequently asked questions
Can employees use free ChatGPT-style tools?
Only for content without personal or confidential data, and only if you allow it. Business accounts with a data processing agreement are the safer default.
Do I need a policy if I have two employees?
Yes, a short one. The rules matter more than the length.
Is this legal advice?
No. These are general templates. Adapt them to your tools and have important policies reviewed.