ChatGPT data protection for small businesses: a proportionate guide for 5 to 100 people
Small firms can protect client data in ChatGPT with five modest steps. Put staff on a business plan, write a one-page policy, train people for twenty minutes, check your UK GDPR duties, and add a simple check that catches mistakes. That is proportionate ChatGPT data protection for a small business of 5 to 100 people. No security team needed. No enterprise contract either.
This guide takes each step in turn. It is written for owners and practice managers, not IT departments.
The real risk at your size
The risk is not hackers. It is ordinary people trying to work faster.
Tom at Ashcombe pastes a client email into ChatGPT to tidy the reply. The email holds a name, an address and a bank sort code. Priya uploads a spreadsheet to summarise it. The sheet lists 300 clients. Nobody meant harm. The data has still left the firm.
Three things decide how serious that is:
- Which account they used. A personal free account is treated very differently from a business plan.
- What they pasted. A rough question is fine. A client list is not.
- Whether you can show what happened. Clients and insurers increasingly ask.
AI is welcome here. The aim is to keep company and client data out of it.
Step 1: choose the right ChatGPT account
This is the cheapest, biggest win. On ChatGPT Free and Plus, OpenAI may use conversations to improve its models unless the user opts out, according to Strac. Business plans work the other way round.
OpenAI's own Enterprise privacy at OpenAI page states: "By default, we do not use your business data for training our models". That covers ChatGPT Business, Enterprise and the API, among others.
ChatGPT Business is the usual fit
ChatGPT Business is the plan once called Team. Tactiq notes OpenAI renamed it on 29 August 2025, with the same features at the time. So older guides that say "Team" mean the same thing.
Per Krater.ai, Standard seats cost $20 per user per month on annual billing, or $25 monthly. Premium seats cost $100 annual or $125 monthly. The minimum is two seats. The maximum is 200 per workspace. That covers every firm this guide is for.
You also get an admin console. AI Subscription Comparison lists user and billing management, role-based access, and custom GPTs kept inside your workspace. When someone leaves, you remove them. Their chats stay with the firm, not on a personal login.
And deleted chats? Jazz Security says they leave OpenAI's systems within 30 days, unless OpenAI must keep them for legal or security reasons.
Enterprise suits larger firms with stricter needs. For a fuller side by side, see ChatGPT Team vs Enterprise vs free.
Rule of thumb: if staff use ChatGPT for work, pay for business seats. Personal free accounts are the gap most small firms leave open.
Step 2: a one-page AI policy
Keep it to one page. Long policies go unread. Here is wording you can adapt.
- Use AI tools for work only through the firm's approved accounts. Currently: ChatGPT Business.
- Do not paste or upload client names, contact details, bank or card details, National Insurance or NHS numbers, or passwords.
- Do not upload whole client files, spreadsheets or lists of people.
- Replace names with stand-ins such as "Client A" before asking.
- Check every answer before it goes to a client. You are responsible for it.
- Ask before trying a new AI tool. Tell us if something goes wrong. You will not be in trouble for reporting it.
Ask each person to agree to it, and keep the date they did. That record matters later.
Step 3: short, practical training
Twenty minutes in a team meeting is enough to start. Cover:
- Which account to use, and how to log in.
- Three real examples from your own work. One safe, one to rewrite, one never to paste.
- How to swap names and numbers for labels before asking.
- Who to tell if something slips through.
Repeat it briefly when someone joins. Refresh it when you approve a new tool.
Step 4: your UK GDPR duties
Size does not exempt you. A ten-person firm has the same core duties as a large one.
The AI Policy Tracker summary of ICO guidance says this. Where AI processing of personal data is likely to be high risk to people, you need a Data Protection Impact Assessment (a written risk check) before you start. The ICO treats most AI involving profiling, large-scale processing or new technology as meeting that bar.
In practice, for a small firm:
- Write down what AI is used for, and what data goes near it.
- If personal data goes in, do the assessment. Keep it short and honest.
- Update your privacy notice so clients know AI is used.
- Check your plan's data processing terms with OpenAI.
If a slip has already happened, read whether pasting a client email into ChatGPT is a breach.
Step 5: a check that catches mistakes
Policy and training cut mistakes. They do not stop them. People are busy. Also, staff use more than ChatGPT: Copilot, Gemini, Claude and others.
A small firm has three realistic options:
| Option | Effort | What it does |
|---|---|---|
| Redact by hand | Relies on each person | Works when people remember |
| Block AI sites | Low | Stops use, and staff find workarounds |
| Check before sending | Low once set up | Takes sensitive data out, so the question still goes |
Helio Guard does the third. It adds a browser extension to Chrome and Edge on Windows PCs. Windows only for now. No Mac. It checks each paste, message and file upload on listed AI sites. Card numbers, bank details and client names are taken out or swapped for stand-ins such as "Client A". The question still goes. Checked on the PC, with fixed rules. We never see what your team types.
Worked example: Jo pastes "Fenwick Joinery, invoice 4471, sort code 20-00-00" into ChatGPT. The sort code is taken out. Fenwick Joinery becomes "Client A". When Jo copies the answer back, the real name returns.
It also keeps each person's dated policy agreement and gives a monthly record for clients or insurers. It is £1 a person a month, and you can see the details on the Helio Guard pricing page.
What a proportionate setup looks like
For most firms of 5 to 100 people: business seats for anyone using ChatGPT, a one-page policy everyone has agreed, twenty minutes of training, a short risk assessment, and a check on every PC. That is a few hours of work and a modest monthly cost. It is also enough to answer a client or insurer honestly.
Common questions about ChatGPT data protection
Is ChatGPT Plus safe enough for a small business?
ChatGPT Plus is a personal plan. Conversations may be used to improve OpenAI's models unless each user opts out, and the firm has no admin control over accounts. ChatGPT Business switches training off by default and gives the firm a workspace it manages, which suits client work much better.
Does a 10-person firm really need a DPIA for ChatGPT?
A 10-person firm needs a Data Protection Impact Assessment if its AI use with personal data is likely to be high risk to people. Firm size does not change that test. If staff never put personal data into ChatGPT, record that decision instead. If they do, a short written assessment is the safe course.
Should a small firm just block ChatGPT?
Blocking ChatGPT is simple but rarely holds. Staff often switch to phones or other AI tools, which leaves less visibility than before. Most small firms do better approving one business account, setting clear rules, and checking what is sent, so people keep the time savings while client data stays out.
Related: AI data loss prevention for small businesses: four options compared on size, setup and cost, UK GDPR and AI tools at work: a plain-English guide to ChatGPT GDPR risk for businesses, AI policy for employees: a UK template you can adapt today